Applied Quantitative

Finance
Using Python for
Financial Analysis

Mauricio Garita
Applied Quantitative Finance
 Mauricio Garita

Applied Quantitative
Finance
Using Python for Financial Analysis
Mauricio Garita
Universidad Francisco Marroquín
Guatemala City, Guatemala

ISBN 978-3-030-29140-2 ISBN 978-3-030-29141-9 (eBook)

https://doi.org/10.1007/978-3-030-29141-9

© The Editor(s) (if applicable) and The Author(s) 2021

This work is subject to copyright. All rights are solely and exclusively licensed by the
Publisher, whether the whole or part of the material is concerned, specifically the rights
of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction
on microfilms or in any other physical way, and transmission or information storage and
retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are
exempt from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and
information in this book are believed to be true and accurate at the date of publication.
Neither the publisher nor the authors or the editors give a warranty, expressed or implied,
with respect to the material contained herein or for any errors or omissions that may have
been made. The publisher remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.

Cover illustration: © Melisa Hasan

This Palgrave Pivot imprint is published by the registered company Springer Nature
Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
I want to thank God, my wife Sonia, Míkel and Maia for being the reason
for keeping moving forward.
 Introduction

This book is aimed for students, professionals, academics and every-

one who wants to learn Python and its application to the stock market.
Therefore, the book begins with the simple implementation of Python,
advances to statistical methods and ultimately reaches the creation of
portfolios.
The book also provides a clear understanding of the different discus-
sions that surge between academic and professional world in the area of
finance. As the reader will see, there are references throughout the book
so that the learning experience may continue after the book is finished.
When considering the different aspects of programming the author
centered on the writing of the code as a whole, so that it can be followed
by the reader. This allows a better learning and an implementation of the
process by the reader, giving a better knowledge about the process, the
errors and how to implement solutions.
As a final note, this book does not aim to be a in-depth finance book
or programming book. This book is written from a practitioner point of
view from finance using programming. The reader should see and ana-
lyze this book from a practitioner perspective, with the purpose of learn-
ing the application of different statistical, mathematical and financial
concepts in the growing language that is Python.
As an author, this book is an effort to close an important gap for
those interested in financial programming and I hope that it opens the
doors to new possibilities.

vii
 What This Book Is and Is Not

This book is not a theoretical book that will explain each and every detail
of each indicator presented. This book centers in applying finance to the
different indicators to offer a hands-on experience.
This book does not cover all the aspects of finance. For example, the
book is centered in technical, quantitative and risk analysis of the stock
market, but it does not cover each and every avenue. This book does not
contain options and futures, Montecarlo Simulations and binomial trees.
The reason is that this book aims to be an introductory to intermediate
level. Considering advanced level there are books far more detailed on
this aspects.
This book does not aim to explain programming language to the
reader. It explains the easiest way to program a portfolio, a MACD, a
VaR and other financial instruments. Also, the code is simple and clean
so that the reader is not overwhelmed by programming. I truly believe
that we can learn to program if we start from the basics and this book
aims for this.
This book does not aim to be perfect. Its aim is to open a discussion
considering finance and the growth of the different methods of the inter-
net. Therefore, the book is explained through Yahoo API because it is
the present live information at every second. Even though, everything
can be elaborated in Excel after uploading the document to Jupyter
notebooks or Google Collaboratory.

ix
x WHAT THIS BOOK IS AND IS NOT

Finally, during the writing of this book, I began reading different

scripts from different authors trying to find the easiest way possible to
understand Python. This book is an effort to make things simple.
Thank you for believing in this book. Thank you for giving it the
opportunity of being in your hands, in your reading, supporting your
future. This book is a tool that I wish I had at the beginning of my
career. Simple enough to understand but solid enough to be reliable.
Sincerely, Mauricio Garita
 Contents

Why Python? 1
Installing Python in the Computer 4
Using Jupyter Notebooks with Python 6
Understanding Jupyter Notebooks 7
Using Google Colab 14
References 17

Learning to Use Python: The Basic Aspects 19

Understanding Numbers in Python 20
Understanding Numbers in Python 21
Using Data Structures in Python 25
What Is a List? 25
How to Create a List? 25
Indexing and Cutting a List 27
Appending Lists 32
Arranging Lists 32
From List to Matrices 33
From List to Dictionaries 33
Modifying a Dictionary 35
Other Interesting Functions of a Dictionary 35
The DataFrame 36
Boolean, Loops and Other Features 42
If, Else and Elif in Python 44

xi
xii CONTENTS

Loops 46
For Loop 46
While Loop 52
List Comprehension 55
References 57

Using FRED® API for Economic Indicators and Data (Example) 59

Installing the FRED® API 59
Using the FRED® API to Retrieve Data 60
First Step 60
Second Step 60
Third Step 60
The Gross Domestic Product 63
The Gross Domestic Product Price Deflator 65
Understanding the Process into the Basics 66
Comparing GDP 67

Using Stock Market Data in Python 71

API Sources 71
Most Important Libraries for Using Data in Python in the Present Book 72
Other Important Libraries Not Used in This Book 74
Suggestion of Libraries for Other Applications 74
Using Python with Yahoo Finance API 75
Using Python with Quandl API 77
Using f.fn( ) for Retreiving Information 79
Using Python with Excel 81
Conclusion Regarding Using Data in Python 83

Statistical Methods Using Python for Analyzing Stocks 85

The Central Limit Theorem 85
Creating a Histogram 88
Creating a Histogram with Line Plots 94
Histograms Using f.fn() 96
Histogram (Percent Change) with Two Variables 96
Histogram (Logarithmic Return) with Two Variables 98
Interquartile Range and Boxplots 99
Boxplot with Two Variables 102
Kernel Density Plot and Volatility 104
 CONTENTS xiii

Kernel Density Plot (Percent Change) with Two Variables 108

Covariance and Correlation 109
Scatterplots and Heatmaps 113
Works Cited 117

Elements for Technical Analysis Using Python 119

The Linear Plot with One Stock Price (Max & Min Values
and the Range) 119
When to Use Linear Plots in Finance 123
The Linear Plot with Two or More Stock Price 123
Linear Plot with Volume 126
Volume of Trade 128
Comparison of Securities with Volume Plots and Closing Prices 129
Candlestick Charts 133
Candlestick Charts and Volume 136
Customizing Candlestick Charts and Volume with **Kwargs 138
OHLC Charts with Volume 138
Line Charts with Volume 140
Moving Average with Matplotlib 142
Moving Average with Mplfinance 147
The Exponential Moving Average (EMA) 149
The Moving Average Convergence Divergence (MACD) with Baseline 153
The Moving Average Convergence Divergence (MACD)
with Signal Line 157
Bollinger Bands ® 160
Backtesting Strategies for Trading 162
Parabolic SAR 162
Fast and Slow Stochastic Oscillators 165
References 169

Valuation and Risk Models with Stocks 171

Creating a Portfolio 171
Calculating Statistical Measures on a Portfolio 174
The Capital Asset Pricing Model 176
The Beta 176
The Beta and the CAPM 182
Sharpe Ratio 188
Traynor Ratio 194
xiv CONTENTS

Jensen’s Measure 197

Information Ratio 201
References 209

Value at Risk 211

Historical VaR(95) 211
Historical VaR(99) 213
VaR for the Next 10 Days 214
Historical Drawdown 218
Wrapping Up the Book—Understanding Performance 222
Portfolio Performance using f.fn() 222
Fund Performance using f.fn() 226
Works Cited 234

Works Cited 235

Index 239
 About the Author

Mauricio Garita began his studies at the Universidad Rafael Landivar

in Guatemala studying Economics and Business Administration. Once he
graduated, he went to Manchester Business School to study a master’s in
international business and to el Instituto de Estudios Bursátiles (IEB) to
study a master’s in asset management. He received his Ph.D. from the
Pontificia Universidad de Salamanca focusing in Sociology and Politics
writing his thesis on the economic impact of the Civil War in Guatemala.
He has worked with the office of the World Bank in Guatemala focus-
ing on financial and economic aspects, he directed the Department of
Business Intelligence in the Secretariat for Central American Economic
Integration (SIECA) and the Department of Business Intelligence of the
Private Sector (CACIF) at Guatemala. He also created the Department
of Academic Relations at the Central American Institute of Fiscal Studies
(ICEFI).
In academics, Mauricio has taught at a masters and bachelors level
courses focused on economics and finance at the Universidad Rafael
Landivar, Universidad del Valle, Universidad del Istmo and lately at
Universidad Francisco Marroquin. He was part of the founders of the
master’s in advanced finance at the Universidad del Valle in Guatemala
and researcher and professor in finance at Universidad Francisco
Marroquin.
On a business level, he founded Simpleconomics which centers on
personal and business finance combined with philosophical views from
the stoic philosophy. He is one of the founders of the Stoic Chapter

xv
xvi ABOUT THE AUTHOR

in Guatemala. He also created a podcast with the name Simpleconomics

which is a podcast focused on finance and philosophy.
His research can be read in Business and Politics, Emerging
Economies and Multinational journals to name a few, in the Routledge
book written in collaboration with John Spillan and Nichols Virzi, in
the Rethinking Taxation in Latin America book by Palgrave Macmillan
and in newspapers such as Prensa Libre of Guatemala and the economic
magazines such as El Economista and Estrategia y Negocios for Central
America.
 List of Figures

Why Python?
Fig. 1 Comparison between R and Python (Source [Pfeiffer 2019]) 3
Fig. 2 Jupyter Notebooks (Source Obtained from the computer
of the author) 8
Fig. 3 Jupyter Notebook—selecting Python (Source Obtained
from the computer of the author) 9
Fig. 4 Creating a folder (Source Obtained from the computer
of the author) 10
Fig. 5 A sample of Jupyter Notebooks (Source Obtained
from the computer of the author) 10
Fig. 6 Installing a package (Source Obtained from the computer
of the author) 12
Fig. 7 Selecting package to install (Source Obtained
from the computer of the author) 12
Fig. 8 Searching for a package (Source Obtained from
the computer of the author) 13
Fig. 9 Process of installing a package (Source Obtained
from the computer of the author) 14
Fig. 10 Creating a new Colab document 15
Fig. 11 Google Colaboratory 16
Fig. 12 Changing name to notebook in Google Colab 16
Fig. 13 Accessing Google Colab documents 16

xvii
xviii LIST OF FIGURES

Using Stock Market Data in Python

Fig. 1 Example of the retrieval of data from Tesla
(Source Obtained from the computer of the author) 77
Fig. 2 Petroleum Prices using Quandl (Source Elaborated
by the author with information from Quandl) 78

Statistical Methods Using Python for Analyzing Stocks

Fig. 1 IBM results using describe 88
Fig. 2 IBM Returns Frequency (Source Elaborated by the author
with information from Yahoo Finance) 90
Fig. 3 IBM histogram with 40 bins (Source Elaborated
by the author with information from Yahoo Finance) 91
Fig. 4 IBM frequency using f.fn() 91
Fig. 5 IBM histogram with logarithmic returns (Source Elaborated
by the author with information from Yahoo Finance) 93
Fig. 6 IBM frequency using f.fn() with logarithmic returns 93
Fig. 7 IBM Returns with axvline (Source Elaborated by the author
with information from Yahoo Finance) 94
Fig. 8 Coca-Cola and Pepsi percent change Histogram
(Source Elaborated by the author with information
from Yahoo Finance) 97
Fig. 9 Coca-Cola and Pepsi logarithmic histogram
(Source Elaborated by the author with information
from Yahoo Finance) 98
Fig. 10 IBM Boxplot (Source Elaborated by the author
with information from Yahoo Finance) 102
Fig. 11 Coca-Cola and Pepsi box-plots (Source Elaborated
by the author with information from Yahoo Finance) 103
Fig. 12 IBM KDE with log returns (Source Elaborated
by the author with information from Yahoo Finance) 107
Fig. 13 Coca-Cola and Pepsi KDE (Source Elaborated
by the author with information from Yahoo Finance) 108
Fig. 14 Scatter Matrix of Nutanix and SP500 (Source Elaborated
by the author with information from Yahoo Finance) 114
Fig. 15 Nutanix and SP500 Scatterplot (Source Elaborated
by the author with information from Yahoo Finance) 115
Fig. 16 Nutanix and SP500 Heatmap (Source Elaborated
by the author with information from Yahoo Finance) 115
 LIST OF FIGURES xix

Elements for Technical Analysis Using Python

Fig. 1 Tesla closing price using Python (Source Elaborated
by the author with information from Yahoo Finance) 121
Fig. 2 Tesla Closing price with different size (Source Elaborated
by the author with information from Yahoo Finance) 121
Fig. 3 Comparison of closing prices in Airline Industry
(Source Elaborated by the author with information
from Yahoo Finance) 125
Fig. 4 Setting the legend with shadow, framealpha, fancybox
and borderpad (Source Elaborated by the author
with information from Yahoo Finance) 126
Fig. 5 Comparison of Volume and Closing price of Tesla
(Source Elaborated by the author with information
from Yahoo Finance) 127
Fig. 6 Total Traded Plot (Source Elaborated by the author
with information from Yahoo Finance) 129
Fig. 7 Volume traded of Tesla, General Motors and Ford
(Source Elaborated by the author with information
from Yahoo Finance) 131
Fig. 8 Closing Price of Ford, General Motors and Tesla
(Source Elaborated by the author with information
from Yahoo Finance) 131
Fig. 9 Comparison of volume and closing price of Ford
(Source Elaborated by the author with information
from Yahoo Finance) 132
Fig. 10 Understanding volume and price with Ford Security
(Source Elaborated by the author with information
from Yahoo Finance) 133
Fig. 11 Candlestick bar representation (Source Created
by Bang [2019]) 134
Fig. 12 Zoom candlestick chart (Source Elaborated by the author
with information from Yahoo Finance) 135
Fig. 13 Dow Jones Candlestick chart with volume
(Source Elaborated by the author with information
from Yahoo Finance) 137
Fig. 14 Candlestick chart and volume chart using colors
(Source Elaborated by the author with information
from Yahoo Finance) 139
Fig. 15 OHLC bars explained (Source From the article
written by Basurto [2020]) 139
xx LIST OF FIGURES

Fig. 16 Dow Jones OHLC chart with volume (Source Elaborated

by the author with information from Yahoo Finance) 141
Fig. 17 Line charts with volume (Source Elaborated by the author
with information from Yahoo Finance) 142
Fig. 18 Simple moving average of Walmart (Source Elaborated
by the author with information from Yahoo Finance) 145
Fig. 19 Simple moving average of Target (Source Elaborated
by the author with information from Yahoo Finance) 146
Fig. 20 Simple moving average of Amazon (Source Elaborated
by the author with information from Yahoo Finance) 147
Fig. 21 Simple moving average for Walmart (Source Elaborated
by the author with information from Yahoo Finance) 148
Fig. 22 Amazon EMA 50, 100, 200 (Source Elaborated
by the author with information from Yahoo Finance) 151
Fig. 23 Target EMA 50, 100,200 (Source Elaborated by the author
with information from Yahoo Finance) 151
Fig. 24 Walmart EMA 50, 100,200 (Source Elaborated
by the author with information from Yahoo Finance) 152
Fig. 25 Amazon MACD with Baseline (Source Elaborated
by the author with information from Yahoo Finance) 155
Fig. 26 Walmart MACD with Baseline (Source Elaborated
by the author with information from Yahoo Finance) 156
Fig. 27 Target MACD with baseline (Source Elaborated
by the author with information from Yahoo Finance) 156
Fig. 28 MACD and Signal Line (Source Elaborated by the author
with information from Yahoo Finance) 159
Fig. 29 Dow Jones Bollinger Bands (Source Elaborated
by the author with information from Yahoo Finance) 162
Fig. 30 AAPL Parabolic SAR 164
Fig. 31 Slow stochastic oscillator 168
Fig. 32 Fast stochastic oscillator 169

Valuation and Risk Models with Stocks

Fig. 1 Cumulative returns of the portfolio (Source Elaborated
by the author with information from Yahoo Finance) 185
Fig. 2 Comparing Benchmark and Portfolio (Source Elaborated
by the author with information from Yahoo Finance) 191
Fig. 3 Correlation plot between Portfolio and Benchmark
(Source Elaborated by the author with information
from Yahoo Finance) 193
 LIST OF FIGURES xxi

Fig. 4 Comparsion between portfolio and benchmark

(Source Elaborated by the author with information
from Yahoo Finance) 197
Fig. 5 Total position of the portfolio (Source Elaborated
by the author with information from Yahoo Finance) 207
Fig. 6 Position behavior on each security (Source Elaborated
by the author with information from Yahoo Finance) 208

Value at Risk
Fig. 1 Cumulative Return of the portfolio (Source Elaborated
by the author with information from Yahoo Finance) 220
Fig. 2 Drawdown of the portfolio (Source Elaborated
by the author with information from Yahoo Finance) 222
Fig. 3 MSAUX SMA 227
Fig. 4 MSAUX EMA 228
Fig. 5 MSAUX Bollinger Bands 229
Fig. 6 MSAUX RSI 230
Fig. 7 Rebase of the closing price in funds and ETF 232
Fig. 8 Funds and ETF montlhy progression 232
 List of Tables

Learning to Use Python: The Basic Aspects

Table 1 Understanding operators 43

Elements for Technical Analysis Using Python

Table 1 Closing time in stock markets 165

Valuation and Risk Models with Stocks

Table 1 Shares outstanding of Tesla, Netflix, Amazon
and Walt Disney 173
Table 2 Market capitalization of Netflix, Tesla, Amazon
and Walt Disney 173
Table 3 Market capitalization and portfolio weight 174
Table 4 The Beta Table 181
Table 5 Investing USD 500,000 183
Table 6 Alpha decision making 201

xxiii
 List of Equations

Statistical Methods Using Python for Analyzing Stocks

Equation 1 Population mean 86
Equation 2 Sample mean 86
Equation 3 Total security return 89
Equation 4 Sturges rule 89
Equation 5 Logarithmic return equation 92
Equation 6 IQR formula 100
Equation 7 Variance formula 104
Equation 8 Standard deviation equation 105
Equation 9 Kernel density estimation—weighting 106
Equation 10 Covariance 110
Equation 11 Correlation 111

Elements for Technical Analysis Using Python

Equation 1 Total Money Traded 128
Equation 2 MACD equation with EMA 154
Equation 3 Uptrend and Downtrend SAR Equation 162
Equation 4 Fast Stochastic Oscillator Equation 165

Valuation and Risk Models with Stocks

Equation 1 Portfolio standard deviation 174
Equation 2 Total risk 176
Equation 3 Beta calculation with correlation 176

xxv
xxvi LIST OF EQUATIONS

Equation 4 Beta calculation with covariance 177

Equation 5 Excess return 182
Equation 6 Capital Asset Pricing Model 182
Equation 7 Portfolio Return 184
Equation 8 Real Risk 186
Equation 9 Beta calculation with covariance 187
Equation 10 Sharpe Ratio 189
Equation 11 Traynor Ratio 194
Equation 12 Jensen’s measure 198

Value at Risk
Equation 1 Value at Risk - position 216
 Why Python?

Abstract This chapter focuses on the importance of Python for the

financial world and its implications in the present and the future of the
financial industry. With the growth in the easiness for accessing infor-
mation, Python becomes relevant in creating faster models for analyz-
ing data. The second objective of this chapter is to understand how to
install Python, analyzing other aspects for analyzing data with Google
Collaboratory without the need of having Python installed in the com-
puter, knowing how to install Anaconda and using Jupyter Notebooks.

Keyword Importance of Python · Jupyter Notebooks · Anaconda ·

Google Collaboratory

When Guido Van Rossum created the languages, he was aiming for a
better, easier language to handle Amoeba which is a microkernel-based
operating system. The creating began in 1989 and in 1991 it was posted
by the author to USENET (Python Organization 2020).
From there, Python has grown into one of the most used program-
ming languages in the world. O’Reilly (2020) has mentioned that in
their online learning environment, Python is preeminent and that it
accounts approximately 10% of the usage in their website. The growth
of the language is demonstrated in the new application programming

© The Author(s), under exclusive license to Springer 1

Nature Switzerland AG 2021
M. Garita, Applied Quantitative Finance,
https://doi.org/10.1007/978-3-030-29141-9_1
2 M. GARITA

interface (API) created by different companies in the field of finance such

as Quantopian, Quandle, Yahoo Finance and Bloomberg to mention a
few.
As mentioned by Bloomberg (2019) the “the lingua franca of finance
is shifting from Microsoft Excel to the open-source programming language
Python”. The use of Python has gone from a desire to a need when
engaging into the world of finance and programming. It has become a
necessity for finance professional despite the background.
The Future Fund—USD140 billion,1 created in 2006 in Australia, is
teaching its investment team Python in order to analyze that more effi-
ciently and rapidly (Burgess and Wells 2020). It has become a must in
the industry and will continue to grow in further years.
Another important aspect considering the growth of Python is the
easiness of its use. The program has a simple and easy to learn syntax, it
is versatile, it is now widely used and has a community that supports the
process of knowledge and upgrading the code (Shaik 2018). Considering
these aspects, the usual question that arises is when choosing between R
and Python.
R was developed by Ross Ihaka and Robert Gentleman in 1992 and
since then it has grown into a widely used language, specifically in the
statistical methods area. The comparison between both languages is
based on the popularity and its use. In 2017, Stack Overflow presented
a survey in which they established that 45% of data scientists use Python
in comparison with just 11.2% using R (Kan 2018). The final decision is
based on the user, but the usual reason is that Python is easy to use and
with the use of Jupyter Notebooks2 it has become extremely helpful con-
sidering data analysis (Fig. 1).
As a conclusion on why use Python for finance, the answer can be
divided into the following categories:

– Ease of use: The language is extremely easy to use and to under-

stand. The process of accessing data from an Excel or an API is
simple. The different commands are intuitive and the website www.
python.org offers an excellent resource for understanding the pro-
cess in the command module.

1 As of February 9th 2020.

2 For more information concerning Jupyter Notebooks visit next chapter.
 WHY PYTHON? 3

Fig. 1 Comparison between R and Python (Source [Pfeiffer 2019])

4 M. GARITA

– Jupyter Notebooks: In the following chapter the use of Jupyter

Notebooks will be discussed deeply, in the meanwhile it is impor-
tant to state that the use of notebooks is helpful in the program-
ming process as well as when elaborating different plots.
– Plots: In the book the Matplotlib library will be used when elabo-
rating plots. Matplotlib is a very useful library for elaborating 2D
plots. The plots are easy to configurate and share.

The main reason for using Python is that it is growing faster than pre-
dicted in the finance sector, giving way to a faster and better analysis of
the stock market. The use of API for gathering information without the
hassle of downloading an excel, getting to know where the information
should be kept or knowing which folder the right folder is to use. The
future of finance is centered on the capacity of the new technology and
Python, for the author, is the best vehicle.

Installing Python in the Computer

The following is based on the installation of Python with the Anaconda
Distribution. For the installation of Python in its original form the user
can visit www.python.org/downloads/ and choose from the differ-
ent options of Operating System. For the purpose of the book the fol-
lowing instructions are for the installation of Python using Anaconda
Distribution.
The process for installing Python is practically the same for macOS
users as for Windows users. Given the scope of the present book, these
are the two operating systems we will be addressing in the installation
of Python. Once installed, the interface of Python does not change con-
cerning the operating system.
The first step is to download Anaconda Navigator. As stated before,
the reason for using Anaconda Navigators is that it simplifies greatly the
process of installing packages and environments needed for Python. The
interface is also useful for other programming languages such as R. It
also provides the possibility of creating uploading information into their
own Cloud with the possibility of sharing it to others.

https://anaconda.org/ by selecting ≪Download Anaconda≫ from the

Anaconda Navigator can be downloaded from their main webpage

menu. This will automatically read the interface of the computer and
guide you to selecting it for macOS, Windows or Linux.
 WHY PYTHON? 5

For this book, it is recommended that the version of Python that

should be downloaded is Python 3.8 version. This version has important
changes in comparison with its last version, the 2.7, mostly in the writing
of the code. Therefore, it could create confusion on the user given that
certain commands and lines of code will be written differently.
The version that is recommended is the 3.7 version which is the one
used in the book. There are certain things that are done differently in the
2.7 version than they are done in the present book, so if the apprentice is
working with a 2.7 version, he must keep this in mind.
When installing the package, the version that downloads by pressing
in the 3.6 version will suffice. There are two other versions as the 64-Bit
Graphical Installer and the 64-Bit Command line installer but for the
purpose of the present book the first one will suffice. Once the package is
downloaded the process is as follows (for Mac OS):

Step 1: The software has to recognize if it is capable of running the

software.
Step 2: Select a destination that is appropriate. The recommendation
is that it is installed in a specific disk.
Step 3: Perform the Standard Installation taking into consideration
the space required. Approximately 2.13 GB.
The installation should be easy and without a problem. If a prob-
lem occurs the Support Community or the support offered by
Anaconda´s team should be of help. The support page can be vis-
ited at the following link: https://www.anaconda.com/support/.

Once the Anaconda Navigator is installed, on the left side a menu can be
read with the categories Home, Environments, Projects (beta), Learning
and Community. For installing a package, it is important to close the
Jupyter Notebook that has been worked on, because if not that package
will not be read by the Notebook.
Once downloaded an installing screen just as the one below will guide
you on the installations.
When installing Anaconda in the Advanced installation options it is recom-
mended to set Python 3.8 as default. This will allow the possibility of sharing
information with other programs which is useful for the rest of the book.
The installation should be followed according to recommendation
and by attending this recommendation you will have the Anaconda
Navigator installed in your computer.
6 M. GARITA

When the user opens the Anaconda Navigator this will lead to the
main screen. In the main screen the user will see on the left side a menu
with the items Home, Environments, Projects (beta), Learning and
Community. More detail on each of the items in the menu will be given
as the book advances.
On the center the user will have a different option such as Jupyterlab,
Jupyter Notebook, qtconsole, spyder, glueviz, orange3, rstudio, vscode.

the exercises. Finally, on the right-hand side the user will see ≪Sign in
The book will be using mostly the Jupyter Notebook environment for

into Anaconda Cloud≫ which gives the possibility of uploading the

notebooks, something that we will explore at the end of the book.
The second step, after installing Anaconda Navigator is to launch and
accommodate our Jupyter Notebooks. By selecting to launch Jupyter
Notebooks, a webpage of the user default search engine will be opened
and on the search bar an address such as localhost:8888/tree will appear.
This is the user's main directory. This is important because:

• In the directory the user Jupyter Notebooks will be stored for

accessing them.
• Any information such as Excel documents or images will be
retrieved from this directory.
• It will be very useful for uploading Jupyter Notebooks to the Cloud.

Using Jupyter Notebooks with Python

The easiness mentioned in the earlier chapter can be made into more
accessible with the use of Jupyter Notebooks. The Jupyter Project began
in 2014 as an open-source software that works perfectly with Python.
Jupyter Notebooks are recommended for data science and in this case
for financial analysis. The main reason for using the Jupyter Notebook
is for the communication between Python and the computer. The con-
junction of both leads to a better approach when coding, visualizing
data, creating kernels and installing updates as well as libraries (365 Data
Science 2020). For installing Jupyter Notebooks in the computer there
are two procedures:

– Installing Jupyter Software directly: In www.jupyter.org/install

there is a simple process to follow the installation for the use of
Jupyter on the user’s computer. In this website it is possible to
install in Windows, Mac and Linux.
 WHY PYTHON? 7

– Installing Anaconda: This is the procedure that is recommended if

the user hasn’t used the command prompt in Windows or its similar
in Mac and Linux. Anaconda can be installed by visiting the web-
page www.anaconda.com/distribution where it can be installed in
Windows, Mac and Linux.

The reason why Anaconda is the recommended procedure is because it is

the easiest way to install packages, manage libraries, analyze data and vis-
ualize results (Anaconda 2020). For the users that are not proficient with
the command prompt or similar in Linux or MacOS, it is recommended
to begin with the Anaconda Distribution.
For the present book, all the code has been written in Jupyter
Notebooks using Anaconda Distribution. The installation of the pack-
ages and the use of the plots with Matplotlib has been simplified by the
use of this distribution system.

Understanding Jupyter Notebooks

The growth of Jupyter Notebooks in the data science field has been
based on the capacity of writing code that can be easily shared and has
a comprehensible order for others to understand and track any changes.
Since its creation in 2014, the multi-language interactive web applica-
tion, is the most useful tool when elaborating financial code (Das 2020)
(Fig. 2).
After installing Anaconda Navigator is to launch and accommodate
our Jupyter Notebooks. By selecting to launch Jupyter Notebooks, a
webpage of your default search engine will be opened and on the search
bar an address such as localhost:8888/tree will appear. This is your main
directory. This is important because:

• In the directory your Jupyter Notebooks will be stored for accessing

them.
• Any information such as Excel documents or images will be
retrieved from this directory.
• It will be very useful for uploading Jupyter Notebooks to the
Cloud.

Once Jupyter Notebooks is open the program leads to the main direc-
tory. This is the directory where Python is installed. In the next segment,
we will discuss how to change the Python directory.
8 M. GARITA

Fig. 2 Jupyter Notebooks (Source Obtained from the computer of the author)

On the top left-hand side there are three important aspects for
Python:

• Files
– Here are the new and the old files. You can create a new file; you
can move files into a new folder or create a new folder. Also, you
can access files that have been created previously.
• Running
– This will give you the information concerning which programs
are being used.
• Clusters
– This will not be a topic for the book since it involves sharing
information with other data science webpages.
On the left-hand side there will be two options:

• Upload
– This will be used to upload a different document, excel or
Jupyter Notebook to the directory.
• New
– Depending on the version and the system, by selecting new you
can build a new Python Notebook, a text file, a folder or prompt
the Terminal.
 WHY PYTHON? 9

Fig. 3 Jupyter Notebook—selecting Python (Source Obtained from the com-

puter of the author)

Once Jupyter Notebook is launched the first step is to create a folder for
the work to be archived. This can be done by selecting the new bar on
the left and then from the options selecting folder (Fig. 3).
Once the folder has been created, it can be renamed by choosing the
folder and then choosing Rename that will appear once you choose the
folder that will be modified (Fig. 4).
Once the folder is created then the python files can be added by
choosing the top left sorting menu and choosing python.
Some important aspects before starting in Jupyter Notebooks (Fig. 5):

• Finding your directory: Type in the bar pwd and press enter. This
will tell where the file is so that it is retrievable, and it can be used
later in the process.
• About the Kernel: A kernel is basically the program that is composed
by the script (what you write on the notebook). It is important to
restart, shutdown or reconnect the kernel depending on the neces-
sity. When running the program, always restart the kernel if you
have been logged out of your computer or been away from the
computer for some time.
10 M. GARITA

Fig. 4 Creating a folder (Source Obtained from the computer of the author)

Fig. 5 A sample of Jupyter Notebooks (Source Obtained from the computer of

the author)

By selecting new and pressing in the Python 3 items, a new window

will be displayed. This is a Jupyter Notebook where all the commands
and information will be implemented. To understand how a Jupyter
Notebook works it is important to define each of the aspects:
 WHY PYTHON? 11

• The name of the notebook

– You will find it on the upper side of the window. When it is new,
the name is usually “Untitled”. By pressing on Untitled you can
change the name to a name more familiar to what you will be
working on.
• The ribbon
– On the ribbon you will find different options. The first one is the
File option which gives the possibility of opening documents,
making a copy, renaming, saving and checkpoint, revert, print
preview, download and Close and Halt. We will see in detail in
the book each of the steps as we continue working.
– The second option will be Edit. It is similar to other programs
such as the Office package where you can change the different
aspects of documents.
– The other options are Insert, Cell, Kernel, Widget and Help. We
will see in detail these aspects since they are relevant to the usage
of Python.

On the center of the Jupyter Notebook you will find the command
prompt line with a In [ ] and a gray line. Here is where the code is writ-
ten and then executed. Once executed you will have an Out [ ] line of
code with the result.
Finally, Jupyter Notebooks are very user-friendly and if a mistake is
made, it leads to a description of the error which allows to know a mis-
take in an easier way and this allows it to be solved.
Installing a package (Fig. 6):
When selecting Environments there are two columns of information.
The first column is titled Search Environments which leads to the dif-
ferent environments that will be used in the process. For example, there
are different environments depending on the use of the environment is
deep learning or Big Data. On the right of the environment there is the
package section which can be sorted by the installed/not installed on the
upper left. If one chose to see the installed packages it will display all the
packages that have been installed in the computer.
On the top right there is a search packages bar that is very useful when
searching for a specific package such as NumPy, a package that is used
consistently throughout the book (Fig. 7).
12 M. GARITA

Fig. 6 Installing a package (Source Obtained from the computer of the author)

Fig. 7 Selecting package to install (Source Obtained from the computer of the
author)
 WHY PYTHON? 13

Fig. 8 Searching for a package (Source Obtained from the computer of the
author)

To install a package is simple using Anaconda Navigator. It can be

more difficult by using the Terminal, this is one of the most important
reasons for using the Anaconda Navigator software. If a package is not
installed there are two choices for searching the package (Fig. 8):

1. Sorting by not install

2. Using the search bar for the specific package.

Once the package is located, the small square box should be selected for
download. On the bottom right of the page is an apply or clear button,
select the apply button for the installation of the package.
By choosing to apply the package a small window will appear that will
search for the package, modify it and solve any problem. Be aware that
this may take some time depending if the package is download and if
there is any information that is needed to download the package. Once
the package is installed it will appear in the installed section and it is
available to be used when installing a library (see next chapter) (Fig. 9).
14 M. GARITA

Fig. 9 Process of installing a package (Source Obtained from the computer of

the author)

Once the packages are installed, the launch Jupyter Notebook can
be launched by going back to Home and pressing on launch under the
Jupyter Notebook symbol. This will open a search engine window and is
the first process to create a notebook.

Using Google Colab

Google Colab or Google Colaboratory is a new interface to execute
Python code created by Google. One of the most important aspects of
using Google Colab is that the notebooks can be shared easily with oth-
ers to collaborate (therefore the name) in the programming project. The
platform is free and it can be accessed by using the Gmail account.
To create a Colab notebook the following process should be followed
(Fig. 10):

Step 1: Create a Drive in https://drive.google.com/.

Step 2: Once the Drive is created, the new tab can be accessed.
 WHY PYTHON? 15

Fig. 10 Creating a new

Colab document

When pressing the New tab to create a new document using the Google
option, one should look for the Google Colaboratory. It is usually in the
more option (Fig. 11).
Once the Google Colaboratory is found, then it can be selected.

Step 3: Name your notebook by pressing beside the ipynb located on

the top right. For this example the name will be changed to example
(Fig. 12).

Once the document is created, the symbol of Google Colaboratory in the

top left can be pressed to return to all the documents that have been cre-
ated (Fig. 13).
16 M. GARITA

Fig. 11 Google
Colaboratory

Fig. 12 Changing name to notebook in Google Colab

Fig. 13 Accessing Google Colab documents

 WHY PYTHON? 17

References
365 Data Science. 2020. Why Python for data science and why Jupyter to code in
Python Articles 11 min read. n.d. Accessed March 2, 2020. https://365datasci-
ence.com/why-python-for-data-science-and-why-jupyter-to-code-in-python/.
Anaconda. 2020. Anaconda distribution. Accessed March 2, 2020. https://
www.anaconda.com/distribution/.
Bloomberg Corporation. 2019. Bloomberg puts the power of Python in hedgers’
hands. 8 March. Accessed March 2, 2020. https://www.bloomberg.com/
professional/blog/bloomberg-puts-power-python-hedgers-hands/..
Burgess, Matthew, and Sarah Wells. 2020. Giant wealth fund seeks managers who
can beat frothy market. 9 February. Accessed March 2, 2020. https://finance.
yahoo.com/news/giant-wealth-fund-seeks-managers-230000386.html.
Das, Sejuti. 2020. Analytics India Magazine. 27 February. Accessed March
17, 2020. https://analyticsindiamag.com/why-jupyter-notebooks-are-so-
popular-among-data-scientists/.
Kan, Chi Nok. 2018. Data Science 101: Is Python better than R? 1 August.
Accessed March 2, 2020. https://towardsdatascience.com/data-science-
101-is-python-better-than-r-b8f258f57b0f.
O’Reilly. 2020. 5 key areas for tech leaders to watch in 2020. 18 February. Accessed
March 2, 2020. https://www.oreilly.com/radar/oreilly-2020-platform-analysis/.
Pfeiffer, Frank. 2019. R versus Python: Which programming language is better for
data science projects in Finance? 28 May. Accessed March 2, 2020. https://
finance-blog.arvato.com/r-versus-python-in-finance/.
Python Organization. 2020. Python organization. 7 January. Accessed March 2,
2020. https://docs.python.org/2/faq/general.html#what-is-python.
Shaik, Naushad. 2018. 5 reasons why learning Python is the best decision. 21
September. Accessed March 2, 2020. https://medium.com/datadrivenin-
vestor/5-reasons-why-i-learned-python-and-why-you-should-learn-it-as-well-
917f781aea05.
 Learning to Use Python: The Basic Aspects

Abstract Python is not a complex language, but it is imperative to

understand how to program different commands based on the necessities
of the user. This chapter focuses on the understanding of numbers, the
use of parenthesis, iterative programming, algebraic operations, manag-
ing data and loops. The chapter aims as a result to help the reader in
understanding the different processes that Python has for managing data,
understanding the importance of each process and how to choose which
one will fit the best model.

Keywords Booleans · Lists · Algebraic operations · Loops

The present chapter is written as the basics for understanding Python

for Finance. If you feel that you are acquainted with the use of python
and the process by which it operates, then move to the next chapter, if
not please elaborate each of the exercises in a Jupyter Notebook because
learning by practicing is the easiest way for learning Python.

© The Author(s), under exclusive license to Springer 19

Nature Switzerland AG 2021
M. Garita, Applied Quantitative Finance,
https://doi.org/10.1007/978-3-030-29141-9_2
20 M. GARITA

Understanding Numbers in Python

To use numbers in Python, the notation is simple, given that an equation
can be computed as simple as follows:

• Entering a Number
– In [1]: 200
– Out [1]: 200

You can achieve this just by pressing enter once the number has been
inserted into the line of command in the Jupyter Notebook.

• Addition in Python
– In [1]: 200 200
– Out [1]: 400
• Subtracting in Python
– In [1]: 200 200
– Out [1]: 0
• Multiplying in Python
– In [1]: 200 200
– Out [1]: 40000
• Dividing in Python
– In [1]: 200 200
– Out [1]: 1.0
• Square root in Python
– In [1]: (−1) (0.5)
– Out [1]: 6.123233995736766e−17+1j

It is important in this aspect to discuss the different numbers that Python

supports:

• Integers: Known as int are whole numbers without decimal points

such as the result of the 200 + 200 = 400. The integers can be long
integers known as long and have unlimited precision and plain inte-
gers that are called integers and have at least a 32 bits precision.
• Floating Real Values: Known as float and it is written with a decimal
point or a scientific notation. An example is the 200/200 = 1.0
• Complex numbers: Known as complex and is a combination of imagi-
nary numbers and real numbers such as the example of the (−1) (
0.5) = 6.123233995736766e−17+1j (Python 2006)
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 21

The properties above can be applied to any number by writing the fol-
lowing code:

• In [1]: int(200 200)

• Out [1]: 1

• In [1]: float(200 200)

• Out [1]: 0.0

• In [1]: complex(200 200)

• Out [1]: 400+0j

This is extremely useful when there is a necessity of changing the num-

bers into a format that is more useful when the necessity arrives.
As seen in the example before, the way to create a square root or to a
power is through the double asterisk (**) just changing the number after
the double asterisk into a decimal or an integer. Asterisk is useful also for
packing arguments into a specific function or to use in positional argu-
ments using keys (Hunner 2018).

• To a power of three
– In [1]: (2) (3)
– Out [1]: 8

• The square root

– In [1]: (4) (0.5)
– Out [1]: 8

Understanding Numbers in Python

During the present chapter, the bitcoin-usd data will be used as an exam-
ple. To add this information, follow the process:

pip install ffn

import ffn

bitcoin = ffn.get('BTC-USD:Close', start=2021-01-01', end='2021-01-01')

bitcoin.tail()
22 M. GARITA

btcusdclose

Date

2021-01-01 29374.152344

2021-01-02 32127.267578

To use numbers in Python, the notation is simple, given that an equa-

tion can be computed as simple as follows:

• Entering a Number
– In [1]: 200
– Out [1]: 200

You can achieve this just by pressing enter once the number has been
inserted into the line of command in the Jupyter Notebook.

• Addition in Python
– In [1]: 200 200
– Out [1]: 400
• Subtracting in Python
– In [1]: 200 200
– Out [1]: 0
• Multiplying in Python
– In [1]: 200 200
– Out [1]: 40000
• Dividing in Python
– In [1]: 200 200
– Out [1]: 1.0
• Square root in Python
– In [1]: (−1) (0.5)
– Out [1]: 6.123233995736766e−17+1j

It is important in this aspect to discuss the different numbers that Python

supports:

• Integers: Known as int are whole numbers without decimal points

such as the result of the 200 + 200 = 400. The integers can be long
integers known as long and have unlimited precision and plain inte-
gers that are called integers and have at least a 32 bits precision.
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 23

• Floating Real Values: Known as float and it is written with a decimal

point or a scientific notation. An example is the 200/200 = 1.0
• Complex numbers: Known as complex and is a combination of
imaginary numbers and real numbers such as the example of the
(−1) (0.5) = 6.123233995736766e−17+1j (Python 2006)

The properties above can be applied to any number by writing the fol-
lowing code:

• In [1]: int(200 200)

• Out [1]: 1

• In [1]: float(200 200)

• Out [1]: 0.0

• In [1]: complex(200 200)

• Out [1]: 400+0j

In the case of bitcoin-usd the numbers are as follow:

btcusdclose

Date
2021-01-01 29374.152344

2021-01-02 32127.267578

Given that the numbers have decimal points, the type of number is a
float. When analyzing using the command type in bitcoin the result
is as follows:
W\SH ELWFRLQ

SDQGDVFRUHIUDPH'DWD)UDPH
24 M. GARITA

This is extremely useful when there is a necessity of changing the num-

bers into a format that is more useful when the necessity arrives.
As seen in the example before, the way to create a square root or to a
power is through the double asterisk (**) just changing the number after
the double asterisk into a decimal or an integer. Asterisk is useful also for
packing arguments into a specific function or to use in positional argu-
ments using keys (Hunner 2018).

• To a power of three
– In [1]: (2) (3)
– Out [1]: 8

• The square root

– In [1]: (4) (0.5)
– Out [1]: 8

The following can be done with the variable bitcoin. Given that there are
two values in the DataFrame (which will be explained in future chapters),
the following process will choose the latest value:
ELWFRLQ>@

btcusdclose
Date

2021-01-02 32127.267578

If the price of the bitcoin in USD is needed to the power of three the
process is as follows:
ELWFRLQ>@ 

btcusdclose
Date

2021-01-02 2.534522e+13

And for the square root:

ELWFRLQ>@ 
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 25

btcusdclose
Date

2021-01-02 171.388892

Using Data Structures in Python

When working with financial data, it is extremely important to know
how to adequate the data based on the process that it will be used. The
purpose of this section is to help understand the most important process
that can be done with data.

What Is a List?
Lists are one of the most important features in Python because it helps in
managing a data frame. When using lists there are certain questions that
are important to answer one by one and it is what will be done in each
chapter.
A list equivalent is an array, they are both objects which contain infor-
mation. There is a difference when comparing a list or an array to a
string, because a string cannot be altered, and it is useful when the infor-
mation in the array will be used to guarantee that the data will be same
during the duration of the program. It is not that useful in finance, since
the data that is aimed for has to change continuously but it is important
to understand the properties that each has.

How to Create a List?

To create a list there are two useful items, (1) the brackets ([ ]) and
(2) the comma (,). Both of them are needed when creating a list first
because the brackets will allow to define the space of the array, meaning
that it will contain the elements. An example below:

• In [1]: my_list = [1,2,3]

• Out [1]: [1,2,3]

There is no limit when adding elements to a list. There is also no restric-

tion when adding to the array a float, a complex or an integer (Python
2020). An example below:
26 M. GARITA

• In [1]: my_list = [1,2,3,0.5,400+0j]

• Out [1]: [1,2,3,0.5,400+0j]

When adding a letter or a word, one has to add the quotes for adding
them to a list. An example below:

• In [1]: my_list = [1,2,3,0.5,400+0j, "apple", "book", "b"]

• Out [1]: [1,2,3,0.5,400+0j, "apple", "book", "b"]

For the following example, two of the most important index will be used
with the following code:

VS\ IIQJHW A*63&&ORVH VWDUW  HQG 

GRZ IIQJHW A'-,&ORVH VWDUW  HQG 

OLVW >VS\>@GRZ>@@

The problem with this process is that given that the information comes
from a DataFrame, the information is not presented properly given the
structure of the information.

gspcclose
Date
2021-01-29 3741.26001, djiclose
Date
2021-01-29 30170.699219]!

Measuring a List
To know how many elements my list has, the function to be used is
len(). Len() will allow us to understand the elements inside the list in
order to create calculations. An example as follows:

• In [1]: my_list = [1,2,3,0.5,400+0j, "apple", "b"]

• Out [1]: [1,2,3,0.5,400+0j, "apple", "b"]

• In [1]: len(my_list)
• Out [1]: 7
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 27

Since the list has been modified to add different elements, the total of
elements is seven (7). For example, in my_list = [10, 10.25, "string"]
there are three elements, if len(my_list) is used the result will be three
(3) because of each element.

In the example using the Dow Jones Industrial Average and

Standard and Poor 500, the same can be applied.

len(list)

The same can be used in a DataFrame, for example in the variable

bitcoin the process can be followed:

len(bitcoin)

Given that the variable bitcoin only had two prices, the length is two (2).

Indexing and Cutting a List

Indexing is very useful for knowing where the data is located and doing
processes with particular data. In finance, this is a very interesting func-
tion because returns can be calculated on a specific date or a simple mov-
ing average based on a specific number. An example as follows:

• In [1]: my_list[1]
• Out [1]: 2

One of the most important aspects when using list is that lists in Python,
as many other computation languages, starts by counting from zero (0).
This is why when indexing the element in position one [1] the result
28 M. GARITA

is the number two, because it is in position one ([1,2,3,0.5,400+0j,

"apple", "b"]).That is one of the most important aspects of using the
function len() because without knowing how many elements are in a list,
it is difficult to understand which position belongs to each element.
In a list there can also be negative indexing, which means that a neg-
ative number can be used for recalling an element in a list. An example
below:

• In [1]: my_list[−1]
• Out [1]: 'b'

If the process is done backward by using a negative number, then the

result is the letter 'b' because as it is the last element [1,2,3,0.5,400+0j,
"apple", ]).Indexing can also be used to recall from a specific num-
ber onward by using the colon (:). An example below:

• In [1]: my_list[1:]
• Out [1]: [2,3,0.5,400+0j, "apple", "b"]).

Before, the example was demonstrated with the Dow Jones and the
S&P 500 by cutting the list as follows to get the last value:
list = [spy[18:],dow[18:]]

The example below takes all the elements except the element in the first
position, in the position of zero (0). It can also be used for indexing
between different values. An example below:

• In [1]: my_list[3:5]
• Out [1]: [0.5,400+0j]

In this case it includes position three (0.5) and the position four (400+j)
but it excludes the word "apple" because it is not included.
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 29

In the Dow Jones and S&P 500 example, the process can be elabo-
rated in the same way as before with the problem of the arrangement:

list = [spy[17:19],dow[17:18]]

[ gspcclose
Date
2021-01-28 3787.379883
2021-01-29 3741.260010, djiclose
Date
2021-01-28 30603.359375]

If the fifth element wants to be included then the most useful process is
to add the sixth position as follows:

• In [1]: my_list[3:6]
• Out [1]: [0.5,400+0j, 'apple']

List can also be used to recall up to a certain element. In this case, the
process is similar to the one above, but the only element used will be
after the colon. An example as follows:

• In [1]: my_list[:5]
• Out [1]: [1,2,3,0.5,400+0j]

As before, with Dow Jones and S&P 500 example, the process can
be elaborated in the same way:

OLVW >VS\>@GRZ>@@

JVSFFORVH
'DWH




GMLFORVH
'DWH




@
30 M. GARITA

Both processes done before can also be done by using negatives, an

example is as follows:

• In [1]: my_list[:−2]
• Out [1]: [1,2,3,0.5,400+0j]

Or:

• In [1]: my_list[−2:]
• Out [1]: ["apple", "b"]

A list can also be added with other list, this is an important feature when
trying to unite two arrays. An example as follows:

• my_list = [1,2,3,0.5,400+0j, "apple", "b"]

• my_list_2 = ["oranges",5,0,5,300+1]

• In [1]: my_list + my_list_2

• Out [1]: [1,2,3,0.5,400+0j, "apple", "b","oranges",5,0,5,300+1]

With just an addition sign (+) two lists were concatenated. This is useful
when created having two sets of data that will be used conjointly.
This can also be used for adding elements:

• In [1]: my_new_list = my_list + my_list_2 + ['Python']

• Out [1]: [1, 2, 3, 0.5, (400+0j), 'apple', 'b', 'oranges', 5, 0.5,
(300+1j), 'Python']

A list can also be duplicated by multiplying it by two:

• In [1]: my_list*2
• Out [1]: [1, 2, 3, 0.5, (400+0j), 'apple', 'b', 1, 2, 3, 0.5, (400+0j),
'apple', 'b']
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 31

In the list created that involved the indexes, the process can be
done as follow for multiplication, with the problem that it even
duplicates the titles in a DataFrame:
OLVW 

JVSFFORVH
'DWH
GMLFORVH
'DWH
JVSFFORVH
'DWH
 GMLFORVH
'DWH
@

If the process is done in the DataFrame by addition, the following

answer will appear as a type error.
OLVW 



7\SH(UURU7UDFHEDFN PRVWUHFHQWFDOO
ODVW
LS\WKRQLQSXWFGGF! LQPRGXOH!
!OLVW

7\SH(UURUFDQRQO\FRQFDWHQDWHOLVW QRWLQW WROLVW

The reason for this type of error is that, as explained before a not
integer given that the information is required from an API, in this
case Yahoo Finance. It could be suggested that for conversion one
should use the int command, but the error is as follow:
7\SH(UURU7UDFHEDFN PRVWUHFHQWFDOO
ODVW
LS\WKRQLQSXWFEFE! LQPRGXOH!
!LQW OLVW

7\SH(UURULQW DUJXPHQWPXVWEHDVWULQJDE\WHVOLNHREMHFWRUD
QXPEHUQRW OLVW
32 M. GARITA

Appending Lists
Another method of adding elements to a list is by the append( ) that is
useful for adding data. Although it can be done by using the addition
sign (+), the append( ) method is used in data structures and it could
behoove the user when using databases.

• In [1]: my_list.append("Python")
• Out [1]: [1, 2, 3, 0.5, (400+0j), 'apple', 'b', 'Python']

Apart from a word, another list of items can be appended into the origi-
nal list. An example below:

• In [1]: my_list.append(my_list_2)
• Out [1]: [1, 2, 3, 0.5, (400+0j), 'apple', 'b', 'Python', ['oranges',
5,0.5, (300+1j)], ['oranges', 5, 0.5, (300+1j)]]

This is useful when creating a new list with data from other sets. In the
example, as follows.

Arranging Lists
A list can be arranged to suit the needs of the user. This is extremely
useful when the list has to be arranged from first to last or it has to be
sorted by a certain requirement. An example as follows with the list
my_list = [1,2,3]

• In [1]: my_list.reverse()
• Out [1]: [3, 2, 1]

Another procedure for arranging list is sorted, which arranges the data
from first to last. This is helpful when the data has to be arranged, for
example, from the smallest return to the highest return. An example as
follows with the list my_list = [4,1,4,3,2,10,20,23]

• In [1]: my_list.sort()
• Out [1]: [1, 2, 3, 4, 4, 10, 20, 23]

An important aspect of this is that the data, once it is sorted this is a

permanent aspect of the list. Therefore, if the data that is being used
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 33

belongs to the original, the best advice is to create a new list with the
original data that can be modified.

From List to Matrices

To nest a list is to arrange data into a matrix. This is an easy procedure
because of the capacity that Python has for modifying data. An exam-
ple below for the understanding of the use of nesting with the following
lists:
a_1 = [1,2,3]
a_2 = [4,5,6]
a_3 = [7,8,9]

• In [1]: matrix = [a_1, a_2, a_3]

• Out [1]: [[1, 2, 3], [4, 5, 6], [7, 8, 9]]

There is certain procedure that can be executed when using matrix that
are similar to a list. An example below:

• In [1]: matrix = [0]

• Out [1]: [[1, 2, 3]

In the example above the data that is being retrieved is the first list. If the
first element of the list will be retrieved the process is as follows:

• In [1]: matrix = [0][0]

• Out [1]: 1

The first bracket is for retrieving the list and the second bracket is for the
first element in the list. This is based on the fact that in Python the posi-
tion zero (0) is the first position.

From List to Dictionaries

A dictionary is an element for handling data that is different to strings
and list. For a dictionary a different programming element is used: the
curly brackets {}. Dictionaries are very flexible, and they are useful when
handling data because it gives a sense of order that is helpful when the
data is from a big set.
34 M. GARITA

A dictionary is divided into two elements: (1) the keys and (2) the
values. The keys are useful because the key will determine the value that
is being requested. The key is separated from the value by a colon (:) to
indicate that the first element is the key and the second element is the
value. An example as follows:

my_dictionary = {'key1': 'value1', 'key2': 'value2', 'key3':'value3'}

• In [1]: my_dictionary['key1']
• Out [1]: 'value1'

A dictionary can handle different types of data such a integers, floats,

complex or long. This is one of the most flexible characteristics that a
dictionary has:

my_dictionary_2 = { }

• In [1]: my_dictionary_2 [ ]
• Out [1]: 23.5

Another important aspect of dictionaries is that it can perform arithmetic

operations with the data contained in the keys.

• In [1]: my_dictionary_2 [ ] − 23
• Out [1]: 0.5

• In [1]: my_dictionary_2 [ ]*10

• Out [1]: 235

• In [1]: my_dictionary_2 [ ] + 23.50

• Out [1]: 47.0

• In [1]: my_dictionary_2 [ ]/2

• Out [1]: 11.75

This is an important feature because it does not alter the data. Each cal-
culation is created apart from the original data, creating results without
any alteration.
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 35

Modifying a Dictionary
There will be moments when it will be necessary to alter the information
inside the dictionaries. It is a really simple process.

• In [1]: my_dictionary_2 [ ] = my_dictionary_2 [ ] + 100

• Out [1]: {'a': 123.5, 'b': 27,982, 'c': 'hola'}

In, the example above 100 was added to the first key and modified the
value for a result of 123.5 This procedure can be done with addition,
multiplication, division, square root or any other arithmetic approach.
An example of this is as follows:

d = {}
d['price'] = '$25000'
d['car'] = 'Peugeot'

• In [1]: d
• Out [1]: {'car': 'Peugeot', 'price': 25000}

If a discount of 10% should be applied to the price of the car the process
would be as follows.

• In [1]: d['price']*(0.9)
• Out [1]: 22500.0

Other Interesting Functions of a Dictionary

As recalled before, the first element of the dictionary, the one that con-
tains the data is retrieved by asking for the key. In this aspect, Python has
the function key to know which are the keys that are included in the dic-
tionary. An Example as follows:

countries = {}
countries = {'Guatemala': 'Guatemala', 'Costa Rica': 'San José',
'Nicaragua': 'Managua'}

• In [1]: countries.keys()
• Out [1]: dict_keys(['Guatemala', 'Costa Rica', 'Nicaragua'])
36 M. GARITA

This allows the user to understand which keys have been chosen and
how to retrieve the data in the list. For example, it can be the name of
the stocks and the values that each stock holds. If the values of the dic-
tionary are not known and want to be retrieved the process is as follows:

• In [1]: countries.keys()
• Out [1]: dict_values(['Guatemala', 'San José', 'Managua'])

The DataFrame
As seen before, the book is centered on creating DataFrames with data
from the internet obtained directly from the webpage. For this, it is
important to understand the use of a DataFrame and the possibilities
that it has when compared to a list.

For the process of understanding a DataFrame, the f.fn() module

will be used and the following stocks:
VWRFNV IIQJHW JPH&ORVHDDSO&ORVHPVIW&ORVH VWDUW  
HQG 
VWRFNVKHDG

gmeclose aaplclose msftclose

Date
2020-01-02 6.31 75.087502 160.619995
2020-01-03 5.88 74.357498 158.619995
2020-01-06 5.85 74.949997 159.029999
2020-01-07 5.52 74.597504 157.580002
2020-01-08 5.72 75.797501 160.089996

To create a DataFrame a library named pandas is needed. In this case

pandas stands for panel data which is basically using data in columns and
rows in order to provide calculations. It also provides a better visualiza-
tion when using data which makes it easier to edit.

Data_Frame =
pd.DataFrame({'col1':[1,2,3,4], 'col2':[200,300,200,300], 'col3':[ 'abc',
'def', 'ghi’, 'xyz']}).
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 37

• In [1]: Data_Frame
• Out [1]:

col1 col2 col3

0 1 200 abc
1 2 300 def
2 3 200 ghi
3 4 300 xyz

A DataFrame has similar properties to a dictionary. For example, the col-

umns can be recalled one by one depending on the information needed.

• In [1]: Data_Frame['col1']
• Out [1]:

0 1
1 2
2 3
3 4
Name: col1, dtype: int64

If there are values repeated such as in the second column, with the
unique command the values can be removed.

• In [1]: Data_Frame['col2'].unique()
• Out [1]: array([200, 300])

In DataFrames, arithmetic functions can also be applied.

• In [1]: Data_Frame['col1'].sum()
• Out [1]: 10

• In [1]: Data_Frame['col1'].mean()
• Out [1]: 2.5

• In [1]: Data_Frame['col1'].multiply(5)
• Out [1]:

0 5
1 10
2 15
3 20
Name: col1, dtype: int64
38 M. GARITA

The DataFrames can be altered to remove a column to create a new

DataFrame with the drop command.

• In [1]: Data_Frame.drop('col1',axis=1)
• Out [1]:

col2 col3
0 200 abc
1 300 def
2 200 ghi
3 300 xyz

By choosing any of the column the specific column is removed in the

DataFrame but the change is not permanent which behooves the user
when trying to undo the process of calculation. DataFrames can also be
sorted just as the dictionaries were sorted:

• In [1]: Data_Frame.sort_values('col2')
• Out [1]:

col1 col2 col3

0 1 200 abc
2 3 200 ghi
1 2 300 def
3 4 300 xyz

DataFrames can also be used with Booleans. This is one of the most
interesting features. An example as follows:

• In [1]: Data_Frame >200

• Out [1]:

col1 col2 col3

0 False False True
1 False True True
2 False False True
3 False True True
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 39

col1 col2 col3

0 1 200 abc
1 2 300 def
2 3 200 ghi
3 4 300 xyz

As a Boolean, the command to search those values that are bigger

than 200. As an interesting aspect, when the comparison is based on let-
ters, the Boolean response is for it to be true, but this is based on com-
paring two elements that are different.
When using the DataFrame with the Yahoo Finance API, there
are important differences that are useful when working with data. For
now, the following exercise is the application of what was demonstrated
before:

– Choosing a column

VWRFNV> JPHFORVH @

'DWH











40 M. GARITA

– Choosing unique values

VWRFNV> DDSOFORVH @XQLTXH

DUUD\ >









– Using sum with a column

VWRFNV> PVIWFORVH @VXP



– Using mean with a column

VWRFNV> PVIWFORVH @PHDQ



– Using multiply with a column

VWRFNV> PVIWFORVH @PXOWLSO\ 

'DWH




 






1DPHPVIWFORVH/HQJWKGW\SHIORDW
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 41

– Using drop to remove a column

VWRFNVGURS PVIWFORVH D[LV 

gmeclose aaplclose
Date
2020-01-02 6.310000 75.087502
2020-01-03 5.880000 74.357498
2020-01-06 5.850000 74.949997
2020-01-07 5.520000 74.597504
2020-01-08 5.720000 75.797501
… … …
2021-01-25 76.790001 142.919998
2021-01-26 147.979996 143.160004
2021–01-27 347.510010 142.059998
2021-01-28 193.600006 137.089996
2021-01-29 316.644714 133.184998

272 rows × 2 columns

– Sort values
VWRFNVVRUWBYDOXHV JPHFORVH

gmeclose aaplclose
Date
2020-04-03 2.800000 60.352501 153.830002
2020-04-02 2.850000 61.232498 155.259995
2020-04-06 3.090000 65.617500 165.270004
2020-04-01 3.250000 60.227501 152.110001
2020-04-07 3.270000 64.857498 163.490005
… … … …
2021-01-25 76.790001 142.919998 229.529999
2021-01-26 147.979996 143.160004 232.330002
2021-01-28 193.600006 137.089996 238.929993
2021-01-29 316.644714 133.184998 234.684998
2021-01-27 347.510010 142.059998 232.899994

272 rows × 3 columns

– Using a Boolean
VWRFNV!
42 M. GARITA

gmeclose aaplclose msftclose

Date
2020-01-02 False False True
2020-01-03 False False True
2020-01-06 False False True
2020-01-07 False False True
2020-01-08 False False True
… … … …
2021-01-25 False True True
2021-01-26 True True True
2021-01-27 True True True
2021-01-28 True False True
2021-01-29 True False True

272 rows × 3 columns

Boolean, Loops and Other Features

Booleans are operators that allow us to know if the element in a function
is TRUE or FALSE. Similar to the example above, the process allowed
us to know which values are bigger than 200. Booleans are extremely
important when analyzing data because they allow for decision-making
in finance.

• In [1]: 1 > 2
• Out [1]: False

• In [1]: 1 < 2
• Out [1]: True

This is one of the most important aspects when comparing data. For
example, if there is a need to know if the Value at Risk (VaR) will be
higher than a specific percent or number, the Boolean could be useful to
understand the process.
To understand how Booleans work, it is important to specify the dif-
ferent cases in which a Boolean creates a response that can be useful for
the user (Table 1).
Booleans can also be used as a chain of comparison. When there are
different elements included, the Booleans can specify if the sequence is
correct. An example below:
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 43

Table 1 Understanding operators

Operator Description Example

== If the values of the operators are equal the result will be 2==1 (FALSE)
TRUE, if not the result will be false 1==1 (TRUE)
!= This operator is the contrary of the (==) sign because 2 != 1 (TRUE)
it means not equal. When using the != the order of the 1 != 1 (FALSE)
exclamation mark (!) and the equal sign (=) has to be as
shown in the example
There is another way of computing the same operator with
the same results, but it is less useful. The use is <>
> This operator specifies that the argument on the left is 2 > 1 (TRUE)
bigger than the argument on the right 1 > 2 (FALSE)
< This operator specifies that the argument on the left is 2 < 1 (FALSE)
smaller than the argument on the right 1 < 2 (TRUE)
<= o >= This operator is known as equal or greater than, or equal 3 >= 3 (TRUE)
or lesser than. This will include the argument that is equal, 3 >= 4 (FALSE)
being the main difference between the greater than and 4 >= 3 (TRUE)
lesser than operators

Source Elaborated by the author

• In [1]: 1 < 2 < 3

• Out [1]: True

Or:

• In [1]: 1 > 10.4 < 3

• Out [1]: False

For Booleans the use of the command (and) or (or) is extremely useful
when comparing equations. This allows the user to understand if various
statement is True or False. An example as follows:

• In [1]: 3 > 2 and 3 < 4

• Out [1]: True

Or:

• In [1]: 3 != 2 and 3 == 2
• Out [1]: False
44 M. GARITA

When using (or), these questions if one statement is correct based on the
other statement. It will tell us if one of the statements is True regarding
if the other statement is False.

• In [1]: 3 < 2 or 1 == 1
• Out [1]: True

Another example:

• In [1]: 3 < 2 or 1 ! = 1
• Out [1]: False

If, Else and Elif in Python

In Python there are three important statements, (1) if, (2) else and (3)
elif. Each of them has a proper way to be used and in which situation it
should be used.

If:
It is used when the user wants there to be an alternative to what is pre-
sented. This is, when the user chooses an action if a case happens.

The Syntax for writing an if statement is as follows:

if example:
performs action1

Example:

• In [1]: if True:
   print ('It is correct')
• Out [1]: It is correct

Else:

The else statement is useful when combined with the if statements. The rea-
son for this is because when we use the if function, we are asking that if
something happens, the program should demonstrate a certain result. With
else, we complement the if function and tell the function to who what hap-
pens if what is commanded is not fulfilled. In this way we can have the two
parts of the equation, what happens and what does not happen.
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 45

Example:

• In [1]: a = False

if a:
print ('a is correct')
else:
print ('anything but correct')

• Out [1]: anything but correct

In this example, what can be observed is that Python understands the True
and the False commands implicitly. These commands that were previously
identified as Boolean, allow the user to establish a relationship. In the pre-
vious case, the variable a was part of the function and the value False placed
a conditional giving as a result that if a is True for it to print a is correct and
the occasion that the result is otherwise the program will print a is anything
but correct. For the value that is True, the following code is as follows:

• In [1]: a = True

if a:
print ('a is correct')
else:
print ('anything but correct')

• Out [1]: a is correct

Elif:

The elif statement is used when there are different arguments. In the
case of an elif statement what the program wants to validate is if the vari-
ables are True or False. An example of an elif statement is as follows:

if expression:
statement(a)
elif expression(2):
statement (a)
elif expression3:
statement (a)
elif expression4:
statement(a)
46 M. GARITA

The following example takes into account the theory of the Value at Risk
(VaR). The purpose of the elif statement is to acknowledge if the fall
of the VaR is one or the other. In this case, the values proposed are in
million. To know more about the VaR, please refer to the chapter titled
Value at Risk (VaR).

In [1]:
var = 100
if var == 200:
print ("1 – Value is True")
print (var)
elif var == 150:
print ("2 – Value is True")
print (var)
elif var == 100:
print ("3 – Value is True")
print (var)
else:
print ("4 – Value is False")
print (var)

Out [1]: 3 – Value is True

100

Loops
There are two types of loops that are extremely useful in Python. The first
is a for loop and the second one is a while loop. Each of the loops has its
own structure and should be used base on the necessity of the program.

For Loop
Loops are iterations that allow the user to elaborate sequences and use
them once and again. For creating a for loop the process although com-
plicated is centered on the purpose of the iteration.
Example:

m = [1,2,3,4,5,6]
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 47

In [1]:
for x in m:
print (x)

Out [1]:
1
2
3
4
5
6

The iteration of the for loop allows us in a simple way to separate each
of the elements included in the list. If a for loop was not used the pro-
cess would have been extremely complicated and also time consuming.
An example of the process without the for loop would have been as
follows:

print (m[0])
print (m[1])
print (m[2])
print (m[3])
print (m[4])
print (m[5])

By using the iteration, it is not only simple but also easy to do a process
with less time. This will be useful when elaborating Monte Carlo simula-
tions or for applying the Sharpe Ratio into a Markowitz model.
Python can also engage in an iteration or a loop without having to
define the element. For example, in the for loop the x was used and it did
not contain an element, this is basic because when the loop ends its itera-
tion this is the out result. If in the loop the x is changed into a word such
as example, the result is the same.

In [1]:
for example in m:
print (example)
48 M. GARITA

Out [1]:
1
2
3
4
5
6

Another interesting thing is that when a string is used with a combina-

tion based on a variable, the result is the replication of the string. This is
useful for understanding the process of a for loop.

m = [1,2,1,2,1]

In [1]:
for x in m:
print ('I am a result')

Out[1]:
I am a result
I am a result
I am a result
I am a result
I am a result

In the example above it is important to understand that it is base on the

elements inside the variable m that amount to five (5) and not on the
numbers that are in the variable [1,2,1,2,1].
Now an important element is the modulo. The modulo allows us to
obtain the waste and is identified by the percent sign (%). We can use this
in the following way.
If the user divides 12 into 5, the result will be that 5 fits twice inside
12 but there is a remainder of 2. In Python the solution can be found in
the following way:

In [1]: 12%5
Out[1]: 2

One of the usual exercises that are practiced when using the mod-
ulo function is the identification of even numbers. Knowing that the
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 49

function modulo (%) gives us the residue, what we must do to find even
numbers is to divide within 2 and the result to be zero (0). An example
as follows:

m = [1,2,3,4,5,6]

In [1]:
for number in m:
if number%2 == 0:
print (number)

Out[1]:
2
4
6

If the example is with odd numbers, the process would only have to
change the part that indicates if number%2 == 0: and change it by if
number%2 == 1:
To the iteration else can be added to operation with a string as a
response. An example is the following:

In [1]:

for number in m:
if number%2 == 0:
print ("the number es even")
else:
print ("the number is uneven")

Out[1]:

the number es even

the number es even
the number es even
the number is uneven

In the same way the arguments can be changed to obtain a specific string
or another element. An example as follows:
50 M. GARITA

In [1]:
for number in m:
if number%2 == 1:
print ("The number is even")
else:
print (number)

Out[1]:

1
The number is even
3
The number is even
5
The number is even

When a loop is written it is important to remember that what is inside

the loop is still being iterated over and over again. Therefore, print is
usually left out of the equation. An example is the following:

m = [1,2,3,4,5,7,8,9,10]

add = 0

In [1]:

for number in m:
add = add + number
print (add)

Out[1]: 49

What the previous interaction does is add the elements of the list m
one after the other to obtain 49. In this case, the for loop is separated
from the print and the determination of the variable because the itera-
tion must be individual, but it will take the other elements.
For loops can also be used for strings with the same modality that are
used for elements.
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 51

In [1]:

a = "Letter"
for letter in a:
print (letter)

Out[1]:
L
e
t
t
e
r

When the process is done in a dictionary there are other elements to

consider. It is important to remember that when making a dictionary the
keys are defined for each of the elements in the list. Therefore, the defi-
nition must be specific for the loops.

In [1]:

dictionary = {'k1':1, 'k2':2, 'k3':3}

for element in dictionary:

print (element)

Out[1]:

k1
k2
k3

To use this in Python 3 the items () command allows us to perform the

function. An example as follows:

In [1]:

for k,v in dictionary.items():

print (k)
print (v)
52 M. GARITA

Out[1]:

k1
1
k2
2
k3
3

When creating a portfolio, the method of a for loop will be used to

group different tickers. The process is as follows:

start = datetime.datetime(2014, 1, 1)
end = datetime.datetime(2019, 1, 1)

tickers = ['NFLX', 'DIS', 'TSLA', 'AMZN']

stocks = pd.DataFrame ()
for x in tickers:
stocks[x] = web.DataReader(x, 'yahoo', start, end)['Close']

While Loop
The while loop will give a result until there is a true value (True). That
is to say that it will be repeated again and again until obtaining the said
result. This can be quite useful to determine a value or a string.

In [1]:

a=0

while a < 10:

print ("a is", a)
a+=1

else:
print ("end")
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 53

Out[1]:

a is 0
a is 1
a is 2
a is 3
a is 4
a is 5
a is 6
a is 7
a is 8
a is 9
end

In this case what is being checked is that the numbers less than 10 will
be added to one (1) until it is less than 10. When it reaches the tenth
(10) value, it will print a phrase that says "end". In this case we can verify
the existence of the elements less than 10 other than being able to add a
data.
To work with while loops it is important to know 3 commands:

1. break,
2. continue and
3. pass

Break allows the user to close the loop that is running. Continue allows
the user to do one more iteration and pass allows us to leave the loop
without effect. While loops can be longer, including more functions to
precisely define what we want.

In [1]:

a=0

while a < 5:
print ("a is", a)
print ("a is less than 5")
a+=1
54 M. GARITA

if a = = 4:
print ("a is equal to 4")
else:
print ("Continue")
continue

Out[1]:

a is 0
a is less than 5
We continue
a is 1
a is less than 5
We continue
a is 2
a is less than 5
We continue
a is 3
a is less than 5
a is equal to 4
a is 4
a is less than 5
continue

What was done in this code is the following.

1. The variable "a" was created and a value of zero (0) was placed.
2. A while loop was created in which when "a" was less than (<) 5
then it would print "a is" and that could be the value of "a". Then
it would print "a is less than 5".
3. An if conditional was placed to say that if the "a" was absolutely
equal (==) to four then it would print "a is equal to 4." For all
the values that are not absolutely equal to four, the string "follow"
would be printed.
4. The "continue" command is placed to give it iteration, that is,
continue replicating.

When we use the break function, we put an end to it. That is why it is
important to use it correctly because it will interrupt the loop.
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 55

In [1]:

a=0

while a < 5:
print ("a is", a)
print ("a is less than 5")
a+=1

if a = = 2:
print ("a is equal to 2")
break
else:
print ("We continue")
continue

Out[1]:

a is 0
a is less than 5
We continue
a is 1
a is less than 5
a is equal to 2

What was indicated in the code is that it stops once it reaches the value
that is absolutely equal to 2. Therefore, when the number 2 is presented,
the last line reads "a is equal to 2" and the code does not continue. This
despite having an "else" that should continue the code.
Finally, if one writes pass result would be the same result as in the loop
without break because it is an argument so that nothing happens.

List Comprehension
The comprehensions of lists allow us to easily and clearly make a list with
a different notation and that this already includes a loop, specifically a for
loop. It is easy to understand as follows:

In [1]:

m = []
56 M. GARITA

for letter in "word":

m.append(letter).
print (m)

Out[1]:

['w', 'o', 'r', 'd'].

In this case it could have been done with a list with the variable "m"
through indexing "word" and that each of the letters was an element.
For example, if one uses m [0] it will give the letter "w" or if we use m
[3] it would give us the letter "d".
The above can be developed in a simple manner as follows:

n = [word for word in "word"]

print (n)

Result

['w', 'o', 'r', 'd']

The lists can also be used for mathematical developments. An example is

that one would like to obtain the square of a list between 0 and 11. For
that, it would be tedious to have to go element by element, developing
it. In this case the simplest form is the following:

In [1]:

list = [x**2 for x in range (0,11)].

print (list).

Out[1]:

[0, 1, 4, 9, 16, 25, 36, 49, 64, 81, 100]

The above can be decompressed as follows. First the variable "list" was
created and assigned to multiply the data within the range of 0 to 11
high (**) squared. In this sense we obtain that 3 by 3 is equal to 9 or
 LEARNING TO USE PYTHON: THE BASIC ASPECTS 57

that 10 by 10 is equal to 100. It is to remember that it reaches number

10 because the ranges do not include the last data.
Previously, it had been analyzed to obtain even results of an equation.
With the lists can be done in a simple way as follows:

In [1]:

list2 = [number for number in range (20) if number%2 == 0]

print (list2)

Out[1]:

[0, 1, 4, 9, 16, 25, 36, 49, 64, 81, 100]

The above is simple because it creates a loop using the function modulo
(%) that indicates the values that have zero residue when being divided
by 2 then they are printed. Since the pairs are the only ones that in a
division of 2 can have zero residue, that is why it is easy to achieve the
above.

References
Hunner, Trey. 2018. Trey Hunner. 11 October. Accessed March 23, 2020.
https://treyhunner.com/2018/10/asterisks-in-python-what-they-
are-and-how-to-use-them/.
Python. 2006. 2.3.4 Numeric types -- int, float, long, complex. 18 October.
Accessed March 23, 2020. https://docs.python.org/2.4/lib/typesnumeric.
html.
Python. 2020. Data structures. 23 March. Accessed March 23, 2020. https://
docs.python.org/3/tutorial/datastructures.html.
Using FRED® API for Economic Indicators
and Data (Example)

Abstract There are different application programming interfaces (API)

for accessing data directly from the web and one of the most important
for economic and financial analysis is the Federal Reserve of Saint Lois
(FRED).

Keywords Data · Inflation · Growth · Deflator

For the application to indicators, first, the API (application programming

interface) of the Federal Reserve of Saint Lois (FRED) will be used to
retrieve information The FRED is an important database for macroeco-
nomic concepts that are necessary for understanding the market. The
example will be developed in Google Colaboratory to exemplify how to
create a notebook that can be shared.

Installing the FRED® API

The FRED® API tool is easy to use and it leads to a faster analysis of the
macroeconomic situation.
For installing the API, one should visit the webpage https://fred.
stlouisfed.org/docs/api/fred/ or google search the term FRED® API to
obtain the result. For the use of the FRED API, it is important to ask
for a key, the key is useful to the FRED because it gives representation

© The Author(s), under exclusive license to Springer 59

Nature Switzerland AG 2021
M. Garita, Applied Quantitative Finance,
https://doi.org/10.1007/978-3-030-29141-9_3
60 M. GARITA

to the person using the API. To obtain the key one should visit the API
Keys page at https://research.stlouisfed.org/docs/api/api_key.html.
The key can be requested by creating a user and filling out the informa-
tion. Once the key is received (it is usually a quick process), then it can
be used for analyzing the data.

Using the FRED® API to Retrieve Data

The first process for using the FRED ® API is to install it in the environ-
ment that will be used. In this example, the Google Collaboratory will
be used given that it is a platform that can be used in similarity with the
use of Anaconda. For installing the FRED® API the process is as follows:

First Step

pip install fredapi

The first process is using pip install1 which allows to retrieve the
information of the FRED. In the present example by using Google
Collaboratory the process is extremely simple.

Second Step
IURP IUHGDSL LPSRUW )UHG
IUHG )UHG DSLBNH\ ;;;;;;;;

In this step, once the fredapi is installed then the Fred data library can
be retrieved. For this it is necessary to have the API_key which will allow
the retrieval of the information.

Third Step
To retrieve information from the FRED one should know how to use
the FRED website and how to retrieve the information. For the follow-
ing exercise will use the information from the FRED website which is as
follows: https://research.stlouisfed.org/.

1 For more information concerning pip install please visit: https://pip.pypa.io/en/

stable/reference/pip_install/.
 USING FRED® API FOR ECONOMIC INDICATORS … 61

The search bar is useful for searching the information which is

needed. In this example the term economic growth will be used, once
entered the option of search only FRED economic data will be chosen.
This filters the result in the data that can be used as an API.
The result page based on the search will return different economic
indicators which can be filtered based on the concepts such as:

– Indexes
– Prices
– Price Index
– Consumer Price Index
– Employment

The information can also be filtered by geography types, geographies,

frequencies, sources, releases and seasonal adjustments. For this example,
the refine of the search won´t be used and the first result on the search
will be the example.
When selected the result Consumer Price Index: Total All Items for the
United States the data is illustrated by the FRED in a line graph. Next
to the name of the Consumer Price Index: Total All Items for the United
States there is a code in parenthesis . This code is useful for
retrieving the information.
It is important to visit the indicator on the webpage to understand the
frequency (in this case it is monthly), the last observation and last update
as well as the units.
In the Google Collaboratory or Anaconda, with the code, it is possible
to retrieve the information. To retrieve the data the process is as follows:
FRQVXPHUBSULFHBLQGH[ IUHGJHWBVHULHV &3$/778601

The variable consumer_price_index has been created and now can be

used for descriptive analysis or to develop an econometric model. The
command tail() can be used to analyze the latest information as follows:
FRQVXPHUBSULFHBLQGH[WDLO






GW\SHIORDW
62 M. GARITA

There is a second method which is easier and works better with the
methods that are going to be seen in this book.

The first step is to install de FRED API:

pip installfredapi

The command pip2 is the program that installs the packages in Python.
Once the fredapi is installed the following process is to install pandas_data-
reader. With pandas_datareader it is possible to access different information
such as FRED, World Bank, OECD and NASDAQ. For more information
concerning the different sources and how to use them please visit: https://
pandas-datareader.readthedocs.io/en/latest/remote_data.html.
LPSRUW SDQGDVBGDWDUHDGHU DV SGU

After pandas_datareader is installed the next part of the process is to install

datetime.3 Datetime is useful for setting dates for retrieving specific data.
Given that the data can be accessed daily, the process is as follows:
VWDUW GDWHWLPHGDWHWLPH 
HQG GDWHWLPHGDWHWLPH 

The start is the date from which the data begins, and the end is where
the data ends. Setting the date is important because it allows the differ-
ent analysis concerning the specific time. With these settings the data can
be retrieved by utilizing the code that was used before.
FRQVXPHUBSULFHBLQGH[ SGU'DWD5HDGHU &3$/778601  IUHG VWDUW
HQG

When the variable is created, the data can be analyzed. To check that
the data is correct, the next process is suggested:

consumer_price_index.tail()
GIBFSL FRQVXPHUBSULFHBLQGH[UHQDPH FROXPQV ^ &3$/778601 
&3, `

2 For more information visit: https://pypi.org/project/pip/.

3 For more information visit: https://docs.python.org/3/library/datetime.html.
 USING FRED® API FOR ECONOMIC INDICATORS … 63

DATE CPALTT01USM657N
2020–05-01 0.001950
2020–06-01 0.547205
2020–07-01 0.505824
2020–08-01 0.315321
2020–09-01 0.139275

What can be observed is that the last data in FRED is from the first
of May 2020. Given that the code could be difficult for
others to understand when sharing the Collaboratory, the title of the col-
umn must be changed:

GIBFSL FRQVXPHUBSULFHBLQGH[UHQDPH FROXPQV ^ &3$/778601 

&3, ` 

The Gross Domestic Product

When analyzing economic cycles, one of the most important variables is
the Gross Domestic Product better known as GDP. The Gross Domestic
Product can be measured through different approaches but the most
useful is the expenditure approach. The expenditure approach is based
on the following formula:
GDP = C + I + G + (X − M)
C = consumption
I = investment
G = Government expenditure
X = Exports
M = Imports

The approach behind the GDP equation is to understand how the econ-
omy is funded and from there understand how it works. The theory is
based that the households offer funds to the banks as savings or to the
companies as investments. Households also offer labor to the companies
in returns of funds, salaries, that can be used for consuming the products
created by the companies or they can be invested and/or saved. Also,
from the salary, the government charges taxes which are used for public
goods and services. Finally, exports and imports are based on the expo-
sure of a country to other economies. In this sense, if a country didn´t
64 M. GARITA

share economic activity with other countries there wouldn´t exist exports
and imports, just consumption, investment and government expenditure.
In order to understand the economic cycle, the GDP will be analyzed.
The Real GDP will be used because it is inflation adjusted and the nom-
inal GDP is established on a base price. The code for the Real GDP is
. The data for the GDP is in billions of dollars, seasonally adjusted
at an annual rate and with a quarterly frequency.

Once the data is retrieved, a change can be made for a more useful
analysis. For this change, the pct.change()4 will be used to analyze the
growth of the GDP. Using pct.change() is simple, when the parenthesis
is left blank, the program assumes that there is one period, which is the
process that will be used.

DATE GDPC1
2019–07-01 0.636915
2019–10-01 0.586232
2020–01-01 –1.262655
2020–04-01 –8.986117
2020–07-01 7.403492

The name of the column can be changed with the process used before.

DATE GDPC1
2019–04-01 0.370716
2019–07-01 0.636915
2019–10-01 0.586232
2020–01-01 –1.262655
2020–04-01 –8.986117
2020–07-01 7.478741
2019–04-01 0.370716

4 For more information please visit: https://pandas.pydata.org/pandas-docs/stable/ref-

erence/api/pandas.DataFrame.pct_change.html.
 USING FRED® API FOR ECONOMIC INDICATORS … 65

The change in the GDP is important to understand the business cycle.

One of the most usual question considering macroeconomics is if the
nominal or real growth should be used when analyzing the GDP.
The nominal GDP reflects the growth without the inflation, meaning
that it considers the output of production by the country and it analyzes
its growth when compared to previous year. Usually, GDP measuring has
a base GDP year which establishes the growth for the nominal GDP. The
main difference is that when using the real GDP is that it reflects the
growth adjusted, leading to the GDP Deflator.

The Gross Domestic Product Price Deflator

The GDP deflator, also known as implicit price deflator for GDP meas-
ures the percentual difference between the nominal GDP and the real
GDP. It is a useful tool for understanding the change in price because it
helps the investor understand if the growth has been motivated by pro-
duction or by prices. The equation is as follows:

nominalGDP
GDPDeflator = × 100
realGDP
To calculate the deflator of the GDP the sum function will be used.
The sum function sums all the data points in the list. The first step is to
establish a base year, in this case the year will be 2017.

VWDUW GDWHWLPHGDWHWLPH 

HQG GDWHWLPHGDWHWLPH 

QRPLQDOB*'3 SGU'DWD5HDGHU > *'3 @ IUHG VWDUWHQG

The comparison will be the year 2019.

VWDUW GDWHWLPHGDWHWLPH 

HQG GDWHWLPHGDWHWLPH 
UHDOB*'3 SGU'DWD5HDGHU > *'3& @ IUHG VWDUWHQG
66 M. GARITA

Once the real GDP and the nominal GDP had been set into a varia-
ble, the following process is to create the sum of the production in the
year.
QJGSBVXP QRPLQDOB*'3VXP
QJGSBVXP LQW QJGSBVXP

UJGSBVXP UHDOB*'3VXP
UJGSBVXP LQW UJGSBVXP

Notice that in the process the variables have been converted to

integers with the function int. The reason for this is that if the data is
divided without converting it into an integer, then the problem is that
Python interprets the data as two different series and cannot unite them
for a process.
With the data converted into integers, then it can be divided.
GHIODWRU QJGSBVXPUJGSBVXP 



The deflator for 2019 in comparison with the nominal GDP in 2017, the
prices have grown by 125%, which reflects that the growth has grown by
boost of prices. Considering that investment it is important to have a growth
higher than inflation, following the deflator is an important indicator.

Understanding the Process into the Basics

One of the most important process is the use of sum instead of the
(+) sign seen in the understanding numbers in Python. Basic arithme-
tic functions can be interchanged with the commands such as sum() or
prod(). These functions are extremely useful when analyzing the data and
developing arithmetic operations in data frames.
When understanding the type of numbers Python supports and reg-
isters, a number such as the variable deflator can be identified with the
command type. One example as follows:

type(deflator)
float
 USING FRED® API FOR ECONOMIC INDICATORS … 67

In the example above the deflator is a float given that it has decimal
numbers. The number could be changed to an integer by using the int()
function.

int(deflator)
102

The variables can be changed and therefore it is important to understand

what each variable represents. For example, if the variables are developed
by a data frame and there is a difficulty when adding them, the variables
can be modified to an integer so that they can be summed. An example
as follows:
GHIODWRU  LQW QRPLQDOB*'3VXP  LQW UHDOB*'3VXP  

The process above changes the sum of the variable nominal_GDP into
an integer so that it can be divided into the sum of the integer of the
real_GDP1 variable.
Considering data structures, given that the information will be
retrieved from an API during this book the result when analyzing list is a
DataFrame.

type(real_GDP)
pandas.core.frame.DataFrame

As mentioned before, a DataFrame is a library in pandas that allow us

to create a more integral analysis. The API used during this book con-
verts the data into a DataFrame automatically and there is not a necessity
to apply the commands learned before.
As seen in the chapter before, this creates an easy access to the func-
tions and on how to work with them. For a better example, let us use
more variables.

Comparing GDP
When analyzing growth, it is important to compare how the growth of
the country that is being analyzed compares to the other countries. In
this process, a graph will be developed to understand how the different
countries behaved.
68 M. GARITA

Installing packages:
LPSRUW SDQGDVBGDWDUHDGHU DV SGU
LPSRUW SDQGDVDV SG
LPSRUW GDWHWLPH
LPSRUW PDWSORWOLES\SORWDV SOW

Setting a date and retrieving information based on the FRED code:

VWDUW GDWHWLPHGDWHWLPH 
HQG GDWHWLPHGDWHWLPH 

JGSBFRPSDULVRQ SGU'DWD5HDGHU > %5$*'314'60(, 

&+1*'314'60(,  86$*'314'60(, @ IUHG VWDUWHQG

Modifying the names on the country list:

JGSBFRPSDULVRQ 
JGSBFRPSDULVRQUHQDPH FROXPQV ^ %5$*'314'60(,  %UD]LO  &+1*'314'60(,
 &KLQD  86$*'314'60(,  86$ `

JGSBFRPSDULVRQKHDG

DATE Brazil China USA

2015–01-01 1.488903e + 12 1.511379e + 13 4.500850e + 12
2015–04-01 1.485916e + 12 1.685497e + 13 4.555894e + 12
2015–07-01 1.504503e + 12 1.765977e + 13 4.586856e + 12
2015–10-01 1.516466e + 12 1.925729e + 13 4.594701e + 12
2016–01-01 1.532747e + 12 1.624100e + 13 4.617539e + 12

Changing the data to percentage change:

JGSBFKDQJH JGSBFRPSDULVRQSFWBFKDQJH  

gdp_change.head()

Dropping NA for creating a chart:

gdp_change = gdp_change.dropna()
gdp_change.head()
 USING FRED® API FOR ECONOMIC INDICATORS … 69

DATE Brazil China USA

2015–04-01 –0.200587 11.520472 1.222980
2015–07-01 1.250849 4.774853 0.679603
2015–10-01 0.795178 9.046097 0.171021
2016–01-01 1.073628 –15.663107 0.497056
2016–04-01 2.034344 11.209285 1.007306

As seen before, this is a DataFrame in which the columns can be

edited, the numbers can be dropped and the information changed. This
is important when applying certain loops as well as when developing a
portfolio, which will be seen in future chapters.
 Using Stock Market Data in Python

Abstract The present chapter aims to demonstrate the different access

to data concerning securities and the packages that are useful to analyze
the data. The book emphasizes Yahoo Finance API, but it explains var-
ious API’s that are accessible for analyzing the data. The packages are
explained and applied to the data.

Keyword API · Packages · Installation · Data

There are different sources of data that are viable in finance. Some of them
can be accessed without creating and CVS (comma-separated-values) file
or an XLS file (Microsoft Excel). The sources that can be accessed online
through Python and without using one of the after mentioned files are the
API (application programming interface). The API is useful because since
they can have a specific protocol, routines and data structures, the infor-
mation accessed can be retrieved every time we are using Python.

API Sources
(1) Google Finance: Google developed an API for retrieving infor-
mation from the financial markets. The API has been discontin-
ued and therefore it is mentioned here as a resource that can be
uploaded in the future.

© The Author(s), under exclusive license to Springer 71

Nature Switzerland AG 2021
M. Garita, Applied Quantitative Finance,
https://doi.org/10.1007/978-3-030-29141-9_4
72 M. GARITA

(2) Yahoo Finance: One of the most important API that the book
will be working with, given that it is free and that the information
is accurate since it is retrieved directly from the markets. Since
Yahoo started its financial platform in 1997 it has grown and its
one of the most consulted platforms for financial decisions that is
accessible by price and by easiness of language.
(3) Quandl: The company's first movement in financial data was
when they launched, in 2013, a million free data set with its
universal API. This created the possibility for analyzing data
from other sources. In late 2018 they were acquired by Nasdaq
(National Association of Securities Dealer), making them one of
the most interesting datasets in the markets.

Although there are other interesting API in the market, in the book we
will be using the two specific API, Yahoo Finance and Quandl. The rea-
son for using these API is because they are gratuitous and that they are
accurate. There are other databases such as Bloomberg that have an API
but access to the data could be expensive.
For the use of CVS and XLS databases, the main databases that will be
used throughout the book are Yahoo Finance for statistical and portfolio
analysis and World Bank and International Monetary Fund for macroe-
conomic investment strategies.

Most Important Libraries for Using Data in Python

in the Present Book

The first API that will be used is the Yahoo Finance API. For this, it is
important to know certain packages that will be important in order to
handle the different data that will be accessed.

(a) NumPy1: The numerical package for Python is one of the most

important packages for handling data. By using NumPy the data
can be approached for linear algebra, multi-dimensional con-
tainers, to create arrays and many other uses. The capability of
NumPy will be demonstrated throughout the book.

1 For more information concerning NumPy: http://www.numpy.org/.

 USING STOCK MARKET DATA … 73

(b) Pandas2: The name of the library is an acronym for Python Data

Analysis Library which is one of the most powerful libraries
when it comes to analyzing data. Pandas can be used to create a
DataFrame, slicing, replacing, creating time series just to name a
few. Pandas will be used throughout the book.
(c) Matplotlib: Throughout the book Matplotlib will be used for
plotting 2D graphs. Matplotlib is an excellent library for plotting
because of its quality, the variety of graphs that can be elaborated
and the easiness of its use.
(d) f.fn( )3: One of the most interesting libraries for quantitative finance
is the f.fn( ) library which helps access data and plot easily. In this
book, the library will be used to access portfolio and measure perfor-
mance. It complements with Matplotlib4 when creating graphs.
(e) Ta-lib: Excellent libraries for developing technical analysis and
backtesting. Will be used with portfolios and stocks. The installa-
tion is tricky, therefore a suggestion is appropriate.

How to install Ta-Lib

The author found it complicated to install Ta-lib. Therefore, it
should be run in the Jupyter Notebook and it cannot be run in the
Google Colab because of the installation. Given that the author
utilizes a Macbook for programming, the process is only stated for
an Apple computer.
First Step: Install Brew in Your Computer.
For installing packages, brew is one of the best in the field. It is
useful because it facilitates the process. For downloading brew fol-
low the present link: https://brew.sh/.
After installing brew it is important to download xcode, which
will help the process of installation. In this case the installation of
xcode it can be done from the Apple Store.
The final process is to install Ta-lib and check that the name of
installation has changed to talib and it is not as described on the
webpage.

2 For more information concerny Pandas: https://pandas.pydata.org/getpandas.html.

3 For more information visit: https://pmorissette.github.io/ffn/index.html.
4 For more information visit: https://ta-lib.org.
74 M. GARITA

To import the other libraries first they must be installed. Please visit the
section on Anaconda Navigator and Installing Packages, this will be very
helpful in the long run.

Other Important Libraries Not Used in This Book

(a) SciPy5: It is a library in Python that uses NumPy and Matplotlib
for mathematics and statistics, including signal processing, linear
algebra, spatial data and special functions to name a few. For more
information
(b) QuantPy6: Although the library is in alpha, this is one of the
libraries that will be interesting to use in the future. The aim is to
simplify calculations used in the present book.
(c) TIA7: It is described as a tool kit to access information, mostly on
the Bloomberg® terminal. Given that the terminal is not easy to
find in many countries, this library is suggested.
(d) PyNance8: was one of the most interesting libraries with f.fn( )
for creating graphs. Sadly part of the code has been deprecated
and it has suffered changes which convert it into complicated for
use and therefore, it is not recommended for the book.

Suggestion of Libraries for Other Applications

Because the purpose of the book is to analyze stocks and learn to use
Python, many of the processes are not created through a specific library,
but by following a step-by-step process and then applying it to a specific
library. Also, when analyzing a library, the author searched for those that
were intuitive for a practitioner moving into Python. The suggestion of
libraries below is part of the next steps for the reader which the author
found interesting but that the above libraries suffice the implementation
with a more accessible framework.

5 For more information visit: https://docs.scipy.org/doc/scipy/reference/.

6 For more information visit: https://github.com/jsmidt/QuantPy.
7 For more information visit: https://github.com/bpsmith/tia.

8 For more information visit: http://pynance.net.

 USING STOCK MARKET DATA … 75

Trading and Backtesting:

– Trade: visit https://github.com/rochars/trade

– Analyzer: visit https://github.com/llazzaro/analyzer
– Finmarketpy: visit https://github.com/cuemacro/finmarketpy

Options:

– Vollib: Visit https://github.com/vollib/vollib

Risk Analysis:

– Empyrical: Visit https://github.com/quantopian/empyrical

– Qfrm: One of the best libraries for bonds. Visit https://pypi.org/
project/qfrm/

Time Series:

– Statsmodels: By far, one of the best libraries for statistical modeling.

Visit https://www.statsmodels.org/stable/index.html

Using Python with Yahoo Finance API

The first step for working with Yahoo API is to establish which libraries
are going to be used. As it was mentioned earlier, NumPy, Pandas and
Matplotlib will be using for accessing the data. The following is the com-
mand on Jupyter Notebooks:

LPSRUW1XP3\DVQS
LPSRUWSDQGDVDVSG
LPSRUWPDWSORWOLES\SORWDVSOW
PDWSORWOLELQOLQH

The PDWSORWOLELQOLQH command is important because it creates a better

graph when elaborating a Jupyter Notebook and also the quality is bet-
ter. It helps the plots to work correctly and it is mentioned as necessary
for the use of Matplotlib.
76 M. GARITA

The second step is to import a module called datetime9 and another

module called pandas_datareader.10 The datetime module is useful for
manipulating dates. This is important given that the information regard-
ing Yahoo API must be determined on a given date. Also, it is useful
when comparing datasets between different stock quotes.
The pandas_datareader allow us to access data from the web. In this
case the pandas_datareader will allow the usage of the API and of the
information.

The script in Jupyter Notebooks is as follows:

LPSRUWSDQGDVBGDWDUHDGHUDVSGU
LPSRUWGDWHWLPH
LPSRUWSDQGDVBGDWDUHDGHUGDWDDVZHE

Once the Jupyter Notebook is set, the Yahoo API can be accessed. The
most important aspect is to set a start date and an end date for the values.
It is also important to add the stock quote that will be consulted, mainly
its ticker. A stock ticker symbol is a one to four letter code representing the
name of the company. For this example, the company Tesla will be used
for analysis. The ticker for Tesla is TSLA. The information is going to be
accessed from January 1, 2015 to January 1, 2019. The code is as follows:

start = datetime.datetime(2015,1,1).
end = datetime.datetime(2019,1,1).

7HVOD ZHE'DWD5HDGHU 76/$  \DKRR VWDUWHQG 

In the example above two variables were created to define the infor-
mation that is going to be accessed. The start date is set to January 1,
2015 and the end date is set to January 1, 2019. Both variables are then
used in the main variable where the Yahoo API is going to be accessed.
The variable Tesla will use the combination of the module web with
DataReader to access the information from the stock company TSLA

9 For more information regarding datetime module: https://docs.python.org/2/library/

datetime.html.
10 For more information regarding pandas_datareader: https://pandas-datareader.readthe-

docs.io/en/latest/.
 USING STOCK MARKET DATA … 77

Fig. 1 Example of the retrieval of data from Tesla (Source Obtained from the
computer of the author)

from the API Yahoo in the dates from January 1, 2015 to January 1,
2019.
Now the information is in the current Jupyter Notebook. It is impor-
tant to recall that every time that we shut down the Jupyter Notebook
the data must be uploaded again using the same procedure or by run-
ning the whole script. To visualize the information the following com-
mand can be executed:

Tesla.tail( ) or Tesla.head( )

The.tail command will show the last five dates of the data. The.head
command will show the first five dates in the beginning of the dates that
were established. Figure 1 demonstrates the capacity of the Yahoo API.
From the Yahoo API the Opening price, the Highest Price of the day the
Lowest Price of the day the Adjusted Close and the Volume are displayed.

Using Python with Quandl API

Quandl is an excellent source of information because it offers databases
for both free and premium. The first step for using Quandl is creating a
user at the signup page at https://www.quandl.com/. This is an impor-
tant aspect because Quandl offers an API Key that is necessary for access-
ing different data.
78 M. GARITA

Fig. 2 Petroleum Prices using Quandl (Source Elaborated by the author with
information from Quandl)

Once the user is created in Quandl, in the account settings it will

appear Your API Key with a lookalike to the following FBWH6W.H\. The key
is personal and it should not be given to other users for access.
With the key, the script in Jupyter is quite easy. The first step will
be to import Quandl, the second step will be to import the API Key
and then create a search based on the Petroleum Prices reported by the
Organization of the Petroleum Exporting Countries (OPEE) from the
following Quandl address: https://www.quandl.com/data/OPEC/
ORB-OPEC-Crude-Oil-Price. The script is as follows:

LPSRUWTXDQGO
TXDQGODSLBFRQILJDSLBNH\  FB7H6W.H\ 
SHWUROHXP TXDQGOJHW 23(&25%DXWKWRNHQ FB[G8/K8N+[Q$NIMD 

Depending on the database there are different aspects that could be

used. For example, the use of the API Key could be inside the variable
because of the need of the program. Another important information
regarding Quandl is that there are databases that are viable when using a
free account such as the OPEC database. An example of the database is
as follows, with the script mentioned above (Fig. 2):
 USING STOCK MARKET DATA … 79

Using f.fn( ) for Retreiving Information

One of the most important libraries for retrieving information that we
will be using in the present book is f.ff( ).11 Certain processes through-
out the book are simplified by the use of this package, the most impor-
tant difficulty is interpreting the results. Therefore, this book will center
on developing a traditional approach and then applying certain packages
for the retrieval of information.
The first step is to install f.fn( ) in the computer. For this exercise
Google Colab will be used:

pip install ffn.

After the installation it is important to import certain packages that

f.fn ( ) uses. The recommended packages are as follows:

Now that it has been installed, the process for retrieving information for
Yahoo Finance is as follows:

As the example above states, the dates are included when getting
the stocks. The stocks are separated by a comma (,) and in the order
they are selected is the order that the stocks will be reflected in the
DataFrame.

stocks.tail()

11 For more information please visit: https://pmorissette.github.io/ffn/ffn.html#ffn.core.

calc_mean_var_weights.
80 M. GARITA

aapl spy amzn fb

Date
2020-12-24 131.970001 369.000000 3172.689941 267.399994
2020-12-28 136.690002 372.170013 3283.959961 277.000000
2020-12-29 134.869995 371.459991 3322.000000 276.779999
2020-12-30 133.720001 371.989990 3285.850098 271.869995
2020-12-31 132.690002 373.880005 3256.929932 273.160004

An important aspect is that f.fn( ) uses adjusted close as a default. If

the interest is to access different prices the process is as follow:

Closing prices:

stocks.tail()

aaplclose spyclose amznclose fbclose

Date
2020-12-24 131.970001 369.000000 3172.689941 267.399994
2020-12-28 136.690002 372.170013 3283.959961 277.000000
2020-12-29 134.869995 371.459991 3322.000000 276.779999
2020-12-30 133.720001 371.989990 3285.850098 271.869995
2020-12-31 132.690002 373.880005 3256.929932 273.160004

High prices:

stocks.tail()

aaplhigh spyhigh amznhigh fbhigh

Date
2020-12-24 133.460007 369.029999 3202.000000 270.399994
2020-12-28 137.339996 372.589996 3304.000000 277.299988
2020-12-29 138.789993 374.000000 3350.649902 280.510010
2020-12-30 135.990005 373.100006 3342.100098 278.079987
2020-12-31 134.740005 374.660004 3282.919922 277.089996
 USING STOCK MARKET DATA … 81

Low prices:

stocks.tail()

aapllow spylow amznlow fblow

Date
2020-12-24 131.100006 367.450012 3169.000000 266.200012
2020-12-28 133.509995 371.070007 3172.689941 265.660004
2020-12-29 134.339996 370.829987 3281.219971 276.279999
2020-12-30 133.399994 371.570007 3282.469971 271.709991
2020-12-31 131.720001 371.230011 3241.199951 269.809998

Using Python with Excel

When using Excel in Python and after the installation of Python using
Anaconda, it is important to know which directory Python is work-
ing on. The following script is necessary for getting to know which
directory Python is working and also on knowing where your files are
located.

Pdw
'/Users/mauriciogarita'

In the above example the directory of the Jupyter Notebook is shown.

The file that is going to be used has to be in the same directory as the
Python in Jupyter Notebook. This is one of the easiest ways of retrieving
the information of an excel. The other option is to know where the doc-
umentation is.
If the documentation is on the same folder or directory as the Jupyter
Notebook that is being used, then the script is as follows:

import pandas as pd.

82 M. GARITA

df

There are other commands such as pd.read_csv12 that can be used

when the file is comma-separated (CSV), or pd.read_table13 that can be
used for a DataFrame that is delimited. Normally, the easiest version of
the pd.read is the pd.read_excel14 because it reads the table, even if it
is not delimited and turns it into a DataFrame that is extremely useful
when handling data.

12 More information concerning the pd.read_csv can be accessed at: https://pandas.

pydata.org/pandas-docs/stable/generated/pandas.read_csv.html.
13 More information concerning the pd.read_table can be accessed at:https://pandas.

pydata.org/pandas-docs/stable/generated/pandas.read_table.html.
14 More information concerning the pd.read_excel can be accessed at: https://pandas.

pydata.org/pandas-docs/stable/generated/pandas.read_excel.html.
 USING STOCK MARKET DATA … 83

Conclusion Regarding Using Data in Python

Python has become a popular language because of the easiness it has
when it comes to handling data. There are different databases that can
be used through an API which includes the information from Quandl,
Google, Yahoo Finance, IEX Finance, Alpha Vantage to name a few.
Using API is the quickest way to access data and to gather different
information depending on the interest on who is accessing the data. It is
also the fastest way to refresh a model and get it up to date.
Of course, that not all the information has an API, for example in a
company that is private. The use of Excel documents is also viable and
fast in Python. As shown in the examples it is easy to access the data
from a file, even if it is a CSV or a table. In the next chapter the informa-
tion will be combined with financial statistics to access data and to create
financial models.
 Statistical Methods Using Python
for Analyzing Stocks

Abstract Statistical Methods are part of the tools for analyzing securi-
ties. The following chapter explains the central limit theorem, returns,
ranges, boxplots, histograms and other sets of statistical measures for the
analysis of securities using Yahoo Finance API.

Keywords Central limit theorem · Returns · Plots · Statistical

measures

This next part of the book is centered on the use of mathematical and
statistical methods to understand the security based on quantitative anal-
ysis. The aim of quantitative analysis is to extract a value that explains
financial behavior (Keaton 2019).

The Central Limit Theorem

The Central Limit Theorem (CLT) is part of the study concerning prob-
ability theory which states that if random samples of a certain size (n)
from any population, the sample will approach a normal distribution. A
normal distribution happens when there is no left or right bias in the
data (Ganti 2019).
The usual representation of a normal distribution is the Bell Curve
which looks lithe a bell, hence the name. CLT establishes that given a

© The Author(s), under exclusive license to Springer 85

Nature Switzerland AG 2021
M. Garita, Applied Quantitative Finance,
https://doi.org/10.1007/978-3-030-29141-9_5
86 M. GARITA

sufficiently large sample size from a population with a specific amount

of variance that is finite, the mean is equal to the median and equal to
the mode. The meaning of this is that there is complete symmetry in the
data and that 50% of the values are higher than the mean and 50% are
lower than the mean.
The mean is usually divided into two: (1) population mean and (2)
sample mean. These aspects are important when analyzing the data that
is going to be used. The formulas give information regarding these
aspects:

Equation 1: Population mean

N

xi
i=1 (1)
µ=
N

Equation 2: Sample mean

n

xi
i (2)
x̄ =
n
N = number of items in the population
n = number of items in the sample

The formulas are basically the same with one important difference,
that the population mean centers on the number of items of a popula-
tion and the sample mean on a specific sample. The population should be
seen as the total observations that can be made and the sample is usually
a part of the population that is selected.

• Importing libraries

import numpy as np
import pandas as pd
import pandas_datareader
import datetime
import pandas_datareader.data as web
import matplotlib.pyplot as plt
%matplotlib inline
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 87

• Setting the continuos data range

start = datetime.datetime(2015,1,1)
end = datetime.datetime(2019,1,1)

• Creating the variable

IBM = web.DataReader(IBM,'yahoo',start,end)

Given that the information used for the IBM security is from January
1st, 2015 to January 1, 2019, this is considered as a sample. Therefore,
the x̄ will be considered as the mean of the security. To calculate the
mean, as the equation suggests, it is the sum of the elements divided by
the count of the elements.

IBM['Close'].mean()

• Result
151.81089374875862

The other aspect mentioned on the CLT is the median. The median
is the middle value of the set of numbers. To calculate the median in
Python it should be calculated as follows:

IBM['Close'].median()

• Result
152.5

Which one to choose? The rule of thumb is to use the mean when
there are no outliers and to use the median when there are outliers.
The third measure of the CLT is the mode. The mode is the most fre-
quent point of data in our data set. In a histogram, it is the highest bar.
To calculate it in Python:

IBM['Close'].mode()

• Result
146.479996
88 M. GARITA

Fig. 1 IBM results

using describe

The process can be combined with the function describe and the pack-
age f.fn () for obtaining certain statistical measures of Central Tendency.
The process is as follows (Fig. 1):

stocks = ffn.get('ibm:Close', start='2015-01-01', end='2019-01-01')

stocks.describe()

Creating a Histogram
One of the most important plots to display information in finance is a
histogram. A histogram demonstrates the frequency of data that is con-
tinuous. The meaning of statistical frequency is the times that a value
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 89

occurs on a set of data. Continuous, refers to continuous data, which

means that the data can take any value within a range.
The histogram for analyzing security has to be elaborated with the
total security return. To calculate the total stock return, the following
formula will be used:

Equation 3: Total security return

(P1 − P0 )
Total Security Return = (3)
P0
P1 = Actual Price
P0 = Previous Price

For example, the company IBM will be used from the before example.
Once the variable IBM is created there are different approaches to which
prices should be used to calculate returns. The conventional approach is
to use closing prices, since the prices are registered as the last price on
the stock. This is useful when working on a historical database.
If there is a necessity for using the actual data, the recommendation is
to use the opening price as the first day and the closing price as the last
day. There is also a discussion concerning the adjusted closing price, but
this will be seen in later chapters.

• Creating the returns

IBM_returns = IBM['Close'].pct_change(1)

The IBM_returns reflects the change in an IBM price. The pct_change is

part of Pandas that gives the percentage of change between the current
and prior number. By establishing the number one (1) it compares the
actual price with the price before.
The IBM_returns can now be converted into a histogram. For this the
matplotlib.pyplot.hist will be used. One of the most important attributes
of hist is the number of bins. The number of bins has as a Rule of Thumb
the Sturge’s Rule.

Equation 4: Sturges rule

K = 1 + 3.322 logN (4)
90 M. GARITA

Fig. 2 IBM Returns Frequency (Source Elaborated by the author with informa-
tion from Yahoo Finance)

K = number of bins
N = number of observations in the sets

To obtain the number of observations the Pandas describe function

can be used as a very useful tool.
IBM_returns.describe()

• Result

count 1006.000000
mean -0.000254
std 0.013023
min -0.076282
25% -0.006450
50% 0.000220
75% 0.006213
max 0.088645
Name: Close, dtype: float64

The total observations are 1006. By including this value, Sturge’s Rule
can be calculated.
bins = 1+3.3222*np.log10(1006)

• Result
10.975231011547681
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 91

Fig. 3 IBM histogram with 40 bins (Source Elaborated by the author with
information from Yahoo Finance)

Fig. 4 IBM frequency using f.fn()

The recommended bins according to Sturge’s Rule is 11. Now the

histogram may be created using 11 bins (Fig. 2).
IBM_returns.hist(bins=11);

It is important to remember that Sturge’s Rule is a rule of thumb. If

the result does not satisfy the criteria, this can be adapted (Fig. 3).
IBM_returns.hist(bins=40);

The same process can be followed with the f.fn() package using a sim-
pler process as follows (Fig. 4):
92 M. GARITA

Calculate percentage returns:

returns = stocks.to_returns().dropna()

Creating a histogram:
returns.hist(bins=40, figsize=(12, 5));

There is another approach for calculating returns and creating a his-

togram. This is through the logarithmic returns.1 The reason for using
logarithmic returns is the normalization of the data, meaning that the
comparison of the data is achievable given that the logarithmic returns
convert the data into a more equal series (Brooks 2008). This assumes
that the distribution is normal, an important aspect of descriptive
statistics.

Equation 5: Logarithmic return equation

 
P1
Logarithmic return = ln (5)
P0

ln = Natural Logarithm
P1 = Actual Price
P0 = Previous Price

To achieve this in Python, it is necessary to use the shift and the np.log.
The shift2 is used to move the calculation by one value, hence the name.
The np.log3 returns the natural logarithm with a base e. Both of them are
necessary to use in order to calculate the logarithmic returns.

IBM_log_returns = np.log(IBM['Close']/IBM['Close'].shift(1))

1 A recommended article on the subject: https://quantivity.wordpress.com/2011/

02/21/why-log-returns/.
2 Pandas documentation on shift: https://pandas.pydata.org/pandas-docs/stable/refer-

ence/api/pandas.DataFrame.shift.html.
3 Numpy documentation on np.log: https://docs.scipy.org/doc/numpy-1.15.0/refer-

ence/generated/numpy.log.html.
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 93

Fig. 5 IBM histogram with logarithmic returns (Source Elaborated by the

author with information from Yahoo Finance)

Fig. 6 IBM frequency using f.fn() with logarithmic returns

When elaborating the comparison of the data, the change between

percentage or logarithmic return is not noticeable, because it is usually in
the fifth decimal point (Fig. 5).
The process can also be elaborated with the f.fn() package by calculat-
ing the logarithmic returns as follows (Fig. 6):

returns = stocks.to_log_returns().dropna()
94 M. GARITA

Fig. 7 IBM Returns with axvline (Source Elaborated by the author with infor-
mation from Yahoo Finance)

After calculating the returns, the process of creating the histogram is similar:

returns.hist(bins=40, figsize=(12, 5));

Creating a Histogram with Line Plots

The above results can be charted into the histogram and the line plot. To add
the results into the histogram an axvline can be added as a part of the plot.
The axvline allows a line to be created to define the plot. In the script, the
mean and the median of the logarithmic returns have been added (Fig. 7).
IBM_log_returns.hist(bins=40);
_ = plt.xlabel('Log Stock Return')

_ = plt.ylabel('Frequency')

_ = plt.title('IBM Log Returns Frequency')

_ = plt.axvline(IBM_log_returns.mean(), color='k', linestyle='dashed', linewidth=1, label =

'mean')

_ = plt.axvline(IBM_log_returns.mean(), color='b', linewidth=0.5, 'median')

plt.legend();
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 95

The plt.axvline was created with the mean function and with the
median function. The color4 was modified and a label was added so
that the legend becomes useful when analyzing the data. Given that the
mean and the median are very similar, there is almost no difference in
the graph but in extreme cases the difference between the mean and the
median can be considerable.
For this, it is necessary to discuss about skewness.5 Since it is important
to understand symmetry, skewness is useful as a measure for understanding
the lack of symmetry. When talking about skewness, the information can
be skewed to the left, to the right or be at the center point (Jain 2018).
IBM_log_returns.skew()

• Result
−0.65950468511581717

The skewness for a normal distribution should be zero and the sym-
metric data should be near this number.

• If the values are positive: data is skewed to the right

• If the values are negative: data is skewed to the left

In the example of IBM, it can be concluded that the data is skewed to

the left and that is near zero which leads to being considered symmetric.
Conjointly with measuring skewness it is important to measure
kurtosis.6 Kurtosis is the measure of the tails in a normal distribution
(Kenton 2019). A high kurtosis is related to having heavy tails, which
means outliers. A low kurtosis means lack of outliers which is light tails.
It is uncommon to have a uniform distribution. To obtain a kurtosis in
Python, the command should be as follows:
IBM_log_returns.kurtosis()

• Result
6.2497976901878483
4 The different colors can be consulted at: https://matplotlib.org/examples/color/

named_colors.html.
5 Pandas information for skewness: https://pandas.pydata.org/pandas-docs/stable/ref-

erence/api/pandas.DataFrame.skew.html.
6 For more information on kurtosis in Pandas: https://pandas.pydata.org/pandas-docs/

stable/reference/api/pandas.DataFrame.kurtosis.html.
96 M. GARITA

There are three options of kurtosis for interpreting the result.

• Leptokurtic: the value of kurtosis is greater than (>) than zero. The
interpretation is that the data is centered around the mean.
• Mesokurtic: the value of the kurtosis is equal (=) to zero. This rep-
resents a normal distribution
• Platykurtic: the value of the kurtosis is less than (<) than zero. The
interpretation is that the data is far from the mean.

In the example of IBM, the kurtosis is leptokurtic, since the values are
around the mean and the shape of the bell is taller in its area.

Histograms Using f.fn()

To create a histogram with one variable with the use of f.fn() the process
is as follows:

Histogram (Percent Change) with Two Variables

The next process is to analyze return by creating a histogram that can be
compared between two companies.

• Installing packages

import numpy as np
import pandas as pd
from pandas_datareader import data as web
import pandas_datareader
import datetime
import matplotlib.pyplot as plt
%matplotlib inline

• Set the dates for the analysis

start = datetime.datetime(2014, 1, 1)
end = datetime.datetime(2019, 1, 1)
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 97

Fig. 8 Coca-Cola and Pepsi percent change Histogram (Source Elaborated by

the author with information from Yahoo Finance)

• Select the companies

CocaCola = web.DataReader('K', 'yahoo', start, end)

Pepsi = web.DataReader('PEP', 'yahoo', start, end)

• Percent Change

Pepsi['Returns'] = Pepsi['Close'].pct_change(1)
CocaCola['Returns'] = CocaCola['Close'].pct_change(1)

One of the columns are created by the append method. The histogram
for both companies can be elaborated as follows (Fig. 8):

• Creating a histogram

CocaCola['Returns'].hist(bins=100,label='Coca-Cola',alpha=0.5)
Pepsi['Returns'].hist(bins=100,label='Pepsi',figsize=(10,8),alpha=0.5)
plt.legend();
98 M. GARITA

Histogram (Logarithmic Return) with Two Variables

The process for creating a comparison with two companies concerning
logarithmic returns is the same as the process for one company. To create
logarithmic returns using the append method, it can be done as follows
(Fig. 9):

• Calculating logarithmic returns

Pepsi['LN Returns'] = np.log(Pepsi['Close']/Pepsi['Close'].shift(1))

CocaCola['LN Returns'] = np.log(CocaCola['Close']/CocaCola['Close'].shift(1))

Fig. 9 Coca-Cola and Pepsi logarithmic histogram (Source Elaborated by the

author with information from Yahoo Finance)
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 99

• Creating a histogram

CocaCola['LN Returns'].hist(bins=100,label='Coca-Cola',alpha=0.5)
Pepsi['LN Returns'].hist(bins=100,label='Pepsi',figsize=(10,8),alpha=0.5)

_ = plt.xlabel('LN Stock Return')

_ = plt.ylabel('Frequency')

_ = plt.title('Coca-Cola & Pepsi - Histogram')

plt.legend();

Interquartile Range and Boxplots

In finance, it is extremely important to measure the spread of data. The
first aspect for measuring the spread of data is the range. The range,
as seen before, is the difference between the highest and lowest values
(Kalla 2020). This is important because it helps in understanding where
the values of the data are located. For calculating the range, the max and
min functions can be used.

IBM_close = IBM['Close']

range_returns = IBM_close.max() - IBM_close.min()

range_returns
• Result

74.379997000000003

Since the max value of the series is 181.94 and the minimum value is
107.57, the difference is the range. This is important to know where the
data is located, between 181.94 and 107.75.
Another important range measure is the Interquartile range (IQR).
The interquartile range is the distance between the third quartile (Q3)
100 M. GARITA

and the first quartile (Q1) (Wan et al. 2014). The Q1 should be under-
stood as one-fourth (25%) of the data is the same or less than the Q1
result. The Q3 should be understood as three-fourths (75%) of the data
is the same or the Q3 result.
The measure is important for understanding outliers. The rule of
thumb is that if a value is larger than Q3 plus 1.5 times the IQR, then
this value is an outlier. This also applies if the value is Q1 minus 1.5
times the IQR range.

Equation 6: IQR formula

IQR = Q3 − Q1 (6)
Outliers
If datapoint > Q3 + 1.5 × IQR

If datapoint < Q1 − 1.5 × IQR

In Python the IQR can be calculated as follows:

IBM_Q1 = IBM_close.quantile(0.25)
IBM_Q1
• Result
144.84000400000002

IBM_Q3 = IBM_close.quantile(0.75)
IBM_Q3
• Result
160.3650055

IBM_IQR = IBM_Q3 - IBM_Q1

IBM_IQR
• Result
15.525001499999973

For the outliers, the process is as follows:

IBM_outlier_high = IBM_Q3 + 1.5 * IBM_IQR

IBM_outlier_high
• Result
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 101

183.65250774999996

IBM_outlier_low = IBM_Q1 - 1.5 * IBM_IQR

IBM_outlier_low
• Result
121.55250175000006

Given the result, there are outliers in the lower range, given that the
minimum value of the is 74.38 and the low outlier is 121.55. There is an
intense discussion about removing outliers, or retaining them consider-
ing that outliers are useful. For the present moment, the outliers should
be kept.
The visualization for analyzing outliers is the boxplots. The boxplots
are also known as the box-whisker plots and they are useful to under-
stand the concentration of the data. To create a boxplot the information
needed is as follows:

• Minimum value outlier

• Q1
• Median
• Q3
• Max value outlier

To create a boxplot in Python (Fig. 10):

plt.boxplot(IBM_close);
_ = plt.ylabel('IBM Closing Price')
_ = plt.title('IBM Close - Boxplot')
_ = plt.xlabel('IBM')

As the boxplot demonstrates, the highest value for measuring an

outlier is 183.65. The beginning of the box below the max value is
Q3 (160.35). The orange line in the middle of the box is the median
(152.5). Below is the Q1 range (144.84) and the last line is the smallest
value for measuring an outlier (121.55). The white and black points are
the outlier of the series, just as calculated by the IQR range and the rule
of thumb for outliers.
102 M. GARITA

Fig. 10 IBM Boxplot (Source Elaborated by the author with information from
Yahoo Finance)

Boxplot with Two Variables

The creation of a Box-Plot is certainly complicated. For instance, both
variables, Pepsi['Close'] and CocaCola['Close'], have to be included in the same
variable. The reason for uniting both variables is that if the process is not
done with concatenation, the ­boxplot will be graphed as one above the
other. To concatenate the pd.concat function of Pandas is extremely use-
ful (Fig. 11).

• Installing packages

import numpy as np
import pandas as pd
from pandas_datareader import data as web
import pandas_datareader
import datetime
import matplotlib.pyplot as plt
%matplotlib inline
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 103

Fig. 11 Coca-Cola and Pepsi box-plots (Source Elaborated by the author with
information from Yahoo Finance)

• Set the dates for the analysis

start = datetime.datetime(2014, 1, 1)
end = datetime.datetime(2019, 1, 1)
104 M. GARITA

• Select the companies

CocaCola = web.DataReader('K', 'yahoo', start, end)

Pepsi = web.DataReader('PEP', 'yahoo', start, end)

• Concatenating

united_box = pd.concat([Pepsi['Close'],CocaCola['Close']],axis=1)

The next step is to create the columns in order to identify clearly the dif-
ference between the prices. The process is as follows:

united_box.columns = ['Pepsi Closing Price','Coca Cola Closing Price']

For plotting, getting the information elaborated with the columns and
combining them for graphs.

united_box.plot(kind='box',figsize=(8,11),colormap='jet')

_ = plt.ylabel('Closing Prices')

_ = plt.title('Pepsi and Coca Cola - Boxplot')

The boxplot demonstrates the outliers concerning Coca-Cola and

Pepsi, as well as the range of the prices. Considering that Pepsi’s closing
prices are higher, this creates an important difference between the height
of the graphs. It is important to consider that the information selected
can be applied to more than two variables.

Kernel Density Plot and Volatility

Although the IQR range is important, there are other elements that
measure the spread of data that are vital in finance: (1) variation and (2)
standard deviation. Variance measures the spread between the numbers
of the data set (Hayes 2019). The formula is as follows:

Equation 7: Variance formula

 (x − x̄)
σ2 = (7)
n
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 105

x = data point
x̄ = mean of the data points in the series
n = total of data points

• Importing libraries

import numpy as np
import pandas as pd
import pandas_datareader
import datetime
import pandas_datareader.data as web
import matplotlib.pyplot as plt
%matplotlib inline

• Setting the continuos data range

start = datetime.datetime(2015,1,1)
end = datetime.datetime(2019,1,1)

• Creating the variable

IBM = web.DataReader(IBM,'yahoo',start,end)

• Calculating the variance

IBM_log_returns.var()
• Result
0.00017102549386496948

The interpretation of the number should be based on a large or small

variance. In finance, a high variance can be led to a riskier asset, mean-
while, a low variance should be interpreted as a low-risk asset. For a
more accurate analysis the standard deviation is used.

Equation 8: Standard deviation equation


 |(x − x̄)|
σ = (8)
n
106 M. GARITA

x = data point
x̄ = mean of the data points in the series
n = total of data points

The standard deviation measures the dispersion of the data set to its
mean. The standard deviation in finance is often referred to as volatility
(Hargrave 2020). Volatility is important because when analyzing higher
volatility, it is usually related to risk. In python it should be computed as
follows:

IBM_log_returns.std()
• Results
0.0130776715765831

Since the standard deviation cannot be negative (because it uses abso-

lute values) and the smallest standard deviation possible is zero, the
standard deviation of IBM logarithmic returns can be seen as a small var-
iation and therefore, a less risky asset. For this it is important to compare
IBM to other companies.
There is a discussion concerning standard deviation, variance and IQR
range. The main comment is that standard deviation and variance are
affected by outliers and therefore, it can be inaccurate. Since IQR range
considers the identification of assets, and it is sometimes seen as a better
measure. Even so, standard deviation and variance are commonly used in
finance.
Finally, one last statistical tool is the Kernel Density Estimation bet-
ter known as the KDE. The KDE plot is usually a replacement for the
histogram because it is calculated by applying a weight to the distance
between points in the dataset. The weighing of distance is achieved by
the following formula:

Equation 9: Kernel density estimation—weighting

  x − observation 
Weighting distance = K (9)
bandwith

K = The kernel function

Bandwidth: distance between observation
x = observations
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 107

Fig. 12 IBM KDE with log returns (Source Elaborated by the author with
information from Yahoo Finance)

The Kernel function can be an Epanechnikov, normal, uniform or tri-

angular. The Pandas data frame kde uses the scott rule7 which is equiv-
alent to the proposal of D.W Scott which establishes an approach to a
normal density.
In conclusion, a KDE is a smoother version of the histogram and can
be compared easily. It is usually graphed with the histogram to compare
the behavior of the variable. In the exercise below it has been plotted
individually (Fig. 12):

IBM_log_returns = np.log(IBM['Close']/IBM['Close'].shift(1))

IBM_log_returns.plot(kind='kde',bw_method='scott', label='IBM',figsize=(12,6));

_ = plt.xlabel('Log Stock Return')

_ = plt.ylabel('Density')

_ = plt.title('IBM Log Returns Frequency')

7 For more information concerning the scott rule: https://docs.scipy.org/doc/scipy/ref-

erence/generated/scipy.stats.gaussian_kde.html#scipy.stats.gaussian_kde.
108 M. GARITA

Kernel Density Plot (Percent Change) with Two

Variables
The elaboration of the KDE plot is simple once the returns have been
created. For this reason, the process is as follows (Fig. 13):

• Installing packages

import numpy as np
import pandas as pd
from pandas_datareader import data as web
import pandas_datareader
import datetime
import matplotlib.pyplot as plt
%matplotlib inline

• Set the dates for the analysis

start = datetime.datetime(2014, 1, 1)
end = datetime.datetime(2019, 1, 1)

Fig. 13 Coca-Cola and Pepsi KDE (Source Elaborated by the author with
information from Yahoo Finance)
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 109

• Select the companies

CocaCola = web.DataReader('K', 'yahoo', start, end)

Pepsi = web.DataReader('PEP', 'yahoo', start, end)

• Calculating returns

Pepsi['Returns'] = Pepsi['Close'].pct_change(1)
CocaCola['Returns'] = CocaCola['Close'].pct_change(1)

• Ellaborating the KDE

CocaCola['Returns'].plot(kind='kde',label='Coca-Cola',figsize=(12,6))
Pepsi['Returns'].plot(kind='kde',label='Pepsi')

_ = plt.xlabel('Stock Return')

_ = plt.ylabel('Density')

_ = plt.title('Coca-Cola & Pepsi - Histogram')

plt.legend();

Covariance and Correlation
When comparing stocks, the technical aspects are important to under-
stand the behavior of the stock when trying to build a portfolio, but they
can be subjective and often misunderstood. There is an interesting dis-
cussion concerning technical analysis and other types of financial analysis
such as fundamental or econometric. It is important to understand the
tool one is using to make the most out of it.
When trying to understand the relation between different asses there
are two quantitative approaches that are very useful: covariance and cor-
relation. Covariance is useful when comparing how two assets are related
(Hayes 2019). It is also an important part of the Capital Asset Pricing
Model (CAPM) that will be explained in future chapters. The equation
of the covariance is as follows:
110 M. GARITA

Equation 10: Covariance

Cov(Ra , Rb ) = E{[Ra − E(Ra )][Rb − E(Rb )]} (10)
To analyze the covariance between two stocks it is important to create
a DataFrame. The process for calculating the covariance is as follows:

• Installing packages

import numpy as np
import pandas as pd
from pandas_datareader import data as web
import pandas_datareader
import datetime
import matplotlib.pyplot as plt
%matplotlib inline

• Establishing dates of the analysis

start = datetime.datetime(2014, 1, 1)
end = datetime.datetime(2019, 1, 1)

• Choosing company and S&P 500

Nutanix = web.DataReader('NTNX', 'yahoo', start, end)

SP500 = web.DataReader('^GSPC', 'yahoo', start, end)

• Calculating returns
Nutanix['Returns'] = Nutanix['Close'].pct_change()
SP500['Returns'] = SP500['Close'].pct_change()

• Remove missing values

Nutanix['Returns'] = Nutanix['Returns'].dropna()
SP500['Returns'] = SP500['Returns'].dropna()

• Create a DataFrame

Nutanix_SP500 = pd.concat([Nutanix['Returns'],SP500['Returns']],axis=1)
Nutanix_SP500.columns = ['Nutanix Returns',' SP500']
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 111

• Co-variance Matrix

covariance = Nutanix_SP500.cov()
covariance

Nutanix Returns SP500

Nutanix Returns 0.001481 0.000119
SP500 0.000119 0.000069

• Annualized Covariance matrix

annual_covariance = Nutanix_SP500.cov() * 252

annual_covariance

Nutanix Returns SP500

Nutanix Returns 0.373307 0.030005
SP500 0.030005 0.017481

To interpret the covariance, one has to analyze if it is positive or nega-

tive, the same way it was done with the variance (Trochim 2020). In this
case the daily covariance is positive but the values are small which means
that they move together but the relation is not strong. When analyzing
the annualized covariance, the same conclusion can be stated. For this it
is important to calculate the correlation.
The correlation demonstrates how strong the relationship is between
two variables, in this case, Nutanix and the S&P500. For determining
correlation, the equation is as follows:

Equation 11: Correlation

Cov(Ra , Rb )
Corr(Ra , Rb ) = (11)
σ (Ra )σ (Rb )
Given that the DataFrame Nutanix_SP500 was created before, the
process is simple.

• Correlation Matrix

correlation = Nutanix_SP500.corr()
correlation
112 M. GARITA

Nutanix Returns SP500

Nutanix Returns 1.0000 0.3919
SP500 0.3919 1.0000

The correlation coefficient ranges from negative (−1.0) to positive

(1.0) (Hayes 2019). When the correlation coefficient is negative (−1.0)
it is said that the relation between variables is perfectly negative, which
means that they behave contrarily. In the situation that the correlation is
positive (1.0) then it is said that they behave perfect positive correlation,
which means that they behave in the same way. The rule of thumb of a
strong correlation to be considered is if the value is 0.8 or above negative
or positive.
In the example of Nutanix, the correlation is positive but it is not
strong, since it is below 0.8. Therefore, there is no correlation to be con-
sidered between Nutanix and the S&P500.
The process can be elaborated with the f.fn() package with certain
simplicity. Therefore it is explained as follows:

First, create a process that includes the stocks with the close price.

stocks = ffn.get('NTNX:Close, spy:Close', start='2014-01-01', end='2019-

01-01')
stocks.tail()

The second step is to create the returns as elaborated in the example

before.

returns = stocks.to_returns()
returns.tail()

ntnxclose spyclose
Date
2018-12-24 −0.009323 −0.026423
2018-12-26 0.077221 0.050525
2018-12-27 0.025437 0.007677
2018-12-28 0.009020 −0.001290
2018-12-31 0.032779 0.008759
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 113

If the returns that are being planned on using are logarithmic returns,
the process is as follows:

returns = stocks.to_log_returns()
returns.tail()

ntnxclose spyclose
Date
2018-12-24 −0.009366 −0.026778
2018-12-26 0.074385 0.049290
2018-12-27 0.025119 0.007648
2018-12-28 0.008980 −0.001291
2018-12-31 0.032253 0.008721

The process will continue with the logarithmic returns. To obtain the
correlation matrix is as follows:

ntnxclose spyclose
ntnxclose 1.000000 0.389009
spyclose 0.389009 1.000000

Scatterplots and Heatmaps
There are two important tools when analyzing the process of comparing
stocks. The first one is the scatterplot which will demonstrate the relation
between variables, in this case the SP500 and Nutanix (Fig. 14).
To create a scatterplot with a histogram and therefore simplify the
process of creating a histogram from scratch pandas has a scatter_matrix
for plotting that can be visualized by using the following command:

• Creating a scatter matrix

from pandas.plotting import scatter_matrix

scatter_matrix(Nutanix_SP500,figsize=(12,10),alpha=1.0,hist_kwds={'bins':90});

The histogram can be adapted by changing the number of bins in the

scatter matrix and the alpha can be changed for more transparency.8 If
8 For more information on scatter_matrix please visit: https://pandas.pydata.org/pan-

das-docs/stable/reference/api/pandas.plotting.scatter_matrix.html.
114 M. GARITA

Fig. 14 Scatter Matrix of Nutanix and SP500 (Source Elaborated by the author
with information from Yahoo Finance)

the only thing needed is the scatterplot, then the process should be as
follows (Fig. 15):

• Creating a scatter plot

Nutanix_SP500.plot(kind='scatter',x='Nutanix Returns',y=' SP500',alpha=0.4,figsize=(15,10));

_ = plt.title('Nutanix and SP500 Scatterplot')

Another interesting way of demonstrating the relation between financial

data is through a heatmap. A heatmap allows to understand the relation
between variables and also scale them considering their relation. The
process uses the seaborn library and is simple. It can be done as follows
(Fig. 16):
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 115

Fig. 15 Nutanix and SP500 Scatterplot (Source Elaborated by the author with
information from Yahoo Finance)

Fig. 16 Nutanix and SP500 Heatmap (Source Elaborated by the author with
information from Yahoo Finance)
116 M. GARITA

• Creating a heatmap

import seaborn as sns

sns.heatmap(correlation,annot=True,cmap=None, linewidths=0.3,annot_kws={"size": 20});

_ = plt.title('Nutanix and SP500 Heatmap')

The heatmap is a very useful tool since the color demonstrates the com-
parison between the different variables and the correlation is expressed
in the boxes. When using different variables, which will the case of the
following chapters, it is important to understand which visual representa-
tion is the most appropriate.
The process can be followed in f.fn() package in the following process:

returns = stocks.to_log_returns()
returns.tail()

ntnxclose spyclose
Date
2018-12-24 −0.009366 −0.026778
2018-12-26 0.074385 0.049290
2018-12-27 0.025119 0.007648
2018-12-28 0.008980 −0.001291
2018-12-31 0.032253 0.008721

returns.corr()

ntnxclose spyclose
ntnxclose 1.000000 0.389009
spyclose 0.389009 1.000000

returns.plot_corr_heatmap();
 STATISTICAL METHODS USING PYTHON FOR ANALYZING STOCKS 117

Works Cited
Brooks, Chris. 2008. Introductory econometrics for finance. Boston: Cambridge
University Press.
Ganti, Akhilesh. 2019. Central Limit Theorem (CLT). 13 September. Accessed
April 2, 2019. https://www.investopedia.com/terms/c/central_limit_theo-
rem.asp.
Hargrave, Marshall. 2020. Standard deviation definition. 1 February. Accessed
February 20, 2020. https://www.investopedia.com/terms/s/standarddevia-
tion.asp.
Hayes, Adam. 2019a. Correlation definition. 20 June. Accessed January 1, 2020.
https://www.investopedia.com/terms/c/correlation.asp.
Hayes, Adam. 2019b. Correlation definition. 20 June. Accessed October 8,
2019. https://www.investopedia.com/terms/c/correlation.asp.
Jain, Diva. 2018. Skew and Kurtosis: 2 Important statistics terms you need to know
in Data Science. 23 August. Accessed August 12, 2019. https://codeburst.
io/2-important-statistics-terms-you-need-to-know-in-data-science-skewness-
and-kurtosis-388fef94eeaa.
Kalla, Siddharth. 2020. Range (Statistics). n.d. Accessed January 4, 2020.
https://explorable.com/range-in-statistics.
118 M. GARITA

Keaton, Will. 2019. Quantitative Analysis (QA). 18 April. Accessed January 15,
2020. https://www.investopedia.com/terms/q/quantitativeanalysis.asp.
Kenton, Will. 2019. Kurtosis. 17 February. Accessed July 30, 2019. https://
www.investopedia.com/terms/k/kurtosis.asp.
Trochim, William M.K. 2020. Correlation. 10 March. Accessed March 12, 2020.
https://conjointly.com/kb/correlation-statistic/.
Wan, Xiang, Wengian Wang, Jiming Liu, and Tiejun Tong. 2014. Estimating
the sample mean and standard deviation from the sample size, median, range
and/or interquartile range. 19 December. Accessed January 03, 2019.
https://bmcmedresmethodol.biomedcentral.com/articles/https://doi.
org/10.1186/1471-2288-14-135.
 Elements for Technical Analysis Using
Python

Abstract Technical Analysis is useful for analyzing the behavior of

securities based on different plots with the purpose of determining the
upward or downward trends. In the present chapter the discussion will
center on returns, volumes, candlestick charts, line charts, simple moving
average, MACD, RSI and other technical analysis tools.

Keywords Central limit theorem · Returns · Plots · Statistical

measures

In the last chapter, the retrieving of data was explored through the pro-
cess of using an API or an Excel file. Once the data is on the Jupyter
Notebook the next step is to proceed with analyzing the information.
The first step is to understand how to display data in Python by using
different methods that it provides.

The Linear Plot with One Stock Price (Max & Min

Values and the Range)
The first plot analyzed in the book is the linear plot. In finance, the
linear plot is useful to understand the trend of the data that is being
used. The linear plot is an excellent visualization tool, although it only

© The Author(s), under exclusive license to Springer 119

Nature Switzerland AG 2021
M. Garita, Applied Quantitative Finance,
https://doi.org/10.1007/978-3-030-29141-9_6
120 M. GARITA

demonstrates how the variable has behaved. In the case of using a stock
price, the linear plot will show how one of the characteristics of the varia-
ble have behaved through a period of time. An example is as follows:

• Importing libraries
LPSRUWQXPS\DVQS
LPSRUWSDQGDVDVSG
LPSRUWSDQGDVBGDWDUHDGHUDVSGU
LPSRUW GDWHWLPH
LPSRUWPDWSORWOLES\SORWDVSOW
PDWSORWOLELQOLQH

• Setting the continuous data range

VWDUW GDWHWLPHGDWHWLPH  
HQG GDWHWLPHGDWHWLPH  

• Creating the variable

7HVOD ZHE'DWD5HDGHU 76/$  \DKRR VWDUWHQG 

Once the variable is created, the plt.plot function is going to be used to

create the variable. The matplotlib.pyplot also has the possibility of creat-
ing a label for the x-axis, a label for the y-axis, and the title for the plot.
To do this, the easiest way is to create a dummy variable by using the
underscore ( _ ). The underscore allows us to add features to a plot with-
out altering the plot. For example (Fig. 1):
B SOWSORW 7HVOD> &ORVH @

B SOW[ODEHO 'DWH 

B SOW\ODEHO &ORVLQJ3ULFH 

B SOWWLWOH 7HVOD&ORVLQJ3ULFH 

There is another way that the linear plot can be elaborated and
according to the author it is easier to work with the problem of the plot
above that is how the Date is shown. The recommended way to build a
linear plot is as follows (Fig. 2):
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 121

Fig. 1 Tesla closing price using Python (Source Elaborated by the author with
information from Yahoo Finance)

Fig. 2 Tesla Closing price with different size (Source Elaborated by the author
with information from Yahoo Finance)

7HVOD> &ORVH @SORW ODEHO 7HVOD ILJVL]H  WLWOH 7HVOD&ORVLQJ3ULFH 

B SOW[ODEHO 'DWH 

B SOW\ODEHO &ORVLQJ3ULFH 
122 M. GARITA

The above chart uses figsize used the width as the first number and the
height in inches. The plt.xlabel and plt.ylabel are used to add the labels
to the x and y-axis. These are attributes that are helpful for creating a
clean-looking linear plot.
When analyzing the data, the lowest closing price seems to be
between January 2016 and July 2017. To find out when the price was
the lowest, argmin can be used to return the minimum value.
Tesla['Close'].argmin()

• Result:

Timestamp('2016-02-10 00:00:00')

As the timestamp demonstrates, the lowest closing price for Tesla was
on the 10th of February. To know the price the stock was trading on this
day:
Tesla['Close'].min()

• Result:

143.66999799999999

The same can be done for the highest closing price traded by Tesla.
Given the chart, it is difficult to analyze when was the highest value pre-
sented. To establish the date argmax can be used on the series:
7HVOD> &ORVH @DUJPD[ 

• Result:

Timestamp('2017-09-18 00:00:00')

To find out what the highest price in the date range, the function max
can be used.
7HVOD> &ORVH @PLQ 

• Result:

385.0
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 123

With the information used by max and by min, we can obtain the
range. The range is important. It tells us the difference between the
highest and the lowest values. The range can be obtained by the differ-
ence between max and min as follows:
7HVOD> &ORVH @PD[ 7HVOD> &ORVH @PLQ 

• Result:

241.33000200000001

When to Use Linear Plots in Finance

Linear plots or line graphs are very useful when analyzing the perfor-
mance of a stock based on its price. Line plots are one of the most useful
resources when representing data such as securities because it allows a
comparison between two or more securities (Halton 2019).
The linear plot gives the possibility of visualizing the security over a
specific period of time, with a reduction in noise based on that they are
elaborated using the closing price (Chen, Line Chart 2019). The use of
the closing price is based on the fact that it is the last price that the secu-
rity has traded at the end of the day in the market.
Using closing prices for security analysis is useful because there is no
alteration in the price once it is closed and it allows a more efficient anal-
ysis. Considering the highest prices of the day, the lowest prices of the
day, the opening price (first price of the day) and the closing price of the
day, the last one is the more commonly used by analysts (Kevin 2015).
An example of two or more stocks is extremely useful for highlighting
the possibility of analysis.

The Linear Plot with Two or More Stock Price

• Importing libraries
124 M. GARITA

LPSRUWQXPS\DVQS
LPSRUWSDQGDVDVSG
LPSRUWSDQGDVBGDWDUHDGHUDVSGU
LPSRUWGDWHWLPH
LPSRUWPDWSORWOLES\SORWDVSOW
PDWSORWOLELQOLQH

• Setting the continuous data range

VWDUW GDWHWLPHGDWHWLPH  

HQG GDWHWLPHGDWHWLPH  

• Creating the variable

$PHULFDQB$LUOLQHV ZHE'DWD5HDGHU $$/  \DKRR VWDUWHQG 

'HOWDB$LUOLQHV ZHE'DWD5HDGHU '$/  \DKRR VWDUWHQG 
8QLWHGB$LUOLQHV ZHE'DWD5HDGHU 8$/  \DKRR VWDUWHQG 
6RXWKZHVWB$LUOLQHV ZHE'DWD5HDGHU /89  \DKRR VWDUWHQG 
-HW%OXB$LUOLQHV ZHE'DWD5HDGHU -%/8  \DKRR VWDUWHQG 

• Creating the chart (Fig. 3)

$PHULFDQB$LUOLQHV> &ORVH @SORW ODEHO $$/ ILJVL]H  WLWOH &ORVLQJ3ULFH 

'HOWDB$LUOLQHV> &ORVH @SORW ODEHO '$/ 
8QLWHGB$LUOLQHV> &ORVH @SORW ODEHO 8$/ 
6RXWKZHVWB$LUOLQHV> &ORVH @SORW ODEHO /89 
-HW%OXB$LUOLQHV> &ORVH @SORW ODEHO -%/8 

B SOW[ODEHO 'DWH 

B SOW\ODEHO &ORVLQJ3ULFH 

SOWOHJHQG 
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 125

Fig. 3 Comparison of closing prices in Airline Industry (Source Elaborated by

the author with information from Yahoo Finance)

One of the main differences with the first linear plot is the use of the
command plt.legend(). The command is useful when there are differ-
ent variables and the user wants to know which line is related to each
security. There are different means of creating a legend, considering the
necessity of the user.

• Examples of different legends

– plt.legend( IDQF\ER[ 7UXH ) adds a shadow around the box.
– plt.legend( IUDPHRQ )DOVHORF ORZHU FHQWHU ) locates the legend in the
lower center of the graph.
– plt.legend( IUDPHRQ )DOVHORF XSSHUFHQWHU ) locates the legend in the
upper center of the graph.
– plt.legend( IUDPHRQ )DOVHORF XSSHU OHIW ) locates the legend in the
upper left of the graph.
– plt.legend( IUDPHRQ )DOVHORF XSSHUULJKW ) locates the legend in the
upper right of the graph.
– plt.legend( IUDPHRQ )DOVHORF XSSHU ULJKW QFRO  ) locates the legend in
the upper right and organizes the different companies in three
columns. The number of columns can be changed.
– plt.legend(
IDQF\ER[ 7UXH IUDPHDOSKD  VKDGRZ )DOVH ERUGHUSDG  ORF ORZHUFHQWHU QFRO 
,ncol=3);
fancybox creates a box around the different items of the legend.
126 M. GARITA

Fig. 4 Setting the legend with shadow, framealpha, fancybox and borderpad
(Source Elaborated by the author with information from Yahoo Finance)

framealpha sets the different shades from white to black. 0 sets the
legend in black and 1 in white.
shadow when it is set to )DOVH eliminated a shadow outside the box
and when it is set to 7UXH it sets a shadow around the box.
borderpad sets the size of the legend box (Fig. 4).

Linear Plot with Volume

The linear plot is not only useful for security prices but it can be also
used for identifying trading volume. The trading volume, better known
in finance as volume, is the number of contracts or shares that are traded
during a period of time (Hayes, Volume Definition 2018). Each trans-
action in the market is quantified by volume, making it one of the most
useful measures of technical analysis. The analysis of the volume has to
be done by comparison with the price of the security. If a higher price is
also followed by a higher volume, the rising of the price is significant.
It is useful to identify if there is a momentum, which can confirm a
trend, or if there is low activity in the market. The example is a conti-
nuity of the Tesla line plot. The variable Tesla that was created has five
categories (1) open, (2) high, (3) low, (4) Close, (5) Adjusted Close and
(6) Volume.

• Creating a Volume Plot (Fig. 5)

 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 127

Fig. 5 Comparison of Volume and Closing price of Tesla (Source Elaborated by

the author with information from Yahoo Finance)

Tesla['Volume'].plot(label='Tesla',figsize=(16,8),title='Volume Traded')

_ = plt.xlabel('Date')
_ = plt.ylabel('Volume')

When compared, the volume seems to have a movement similar to the

closing price of the stock market, assuring that the move is significant.
128 M. GARITA

When the price has been higher there has been more movement in the
stocks concerning Tesla.

Volume of Trade
One of the most interesting aspects when analyzing volume is to mul-
tiply it by the security price, giving as a result the volume of trade dur-
ing a specific period in terms of money invested. The total money traded
is useful to know the quantity of investment in a single day or period,
which is useful when analyzing if the market is selling or buying specific
security, or to understand the monetary impact of trade. The equation is
rather simple:

Equation 1: Total Money Traded

Total Money Trade = Volume × Security Price (1)
• Obtaining Total Money Traded
Tesla_total_traded = Tesla['Close'] * Tesla['Volume']

• Plotting the Total Money Traded (Fig. 6)

Tesla_total_traded.plot(label='Tesla',figsize=(16,8),title='Total Traded')

_ = plt.xlabel('Date')

_ = plt.ylabel('Total Traded')

The Total Money Traded Plot demonstrates a similar tendency to the

Volume plot, but it also identifies was higher. Meaning that the Total
Traded Plot can be seen as the real growth since it includes the volume of
transactions and the current price of the transactions. The Total Traded
Plot is better used when compared with other companies.
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 129

Fig. 6 Total Traded Plot (Source Elaborated by the author with information
from Yahoo Finance)

Comparison of Securities with Volume Plots

and Closing Prices

Comparing volume between securities is vital for financial analysis, basi-

cally because it allows the trader to understand trading confirmation,
exhaust moves and volume, bullish signs to name a few (Mitchell, How
to Use Volume to Improve Your Trading 2020).
To create a plot comparing volumes between securities, the process is
as follows:

• Importing libraries

LPSRUWQXPS\DVQS
LPSRUWSDQGDVDVSG
LPSRUWSDQGDVBGDWDUHDGHUDVSGU
LPSRUWGDWHWLPH
LPSRUWPDWSORWOLES\SORWDVSOW
PDWSORWOLELQOLQH
130 M. GARITA

• Setting the continuous data range

VWDUW GDWHWLPHGDWHWLPH  

HQG GDWHWLPHGDWHWLPH  

• Choosing securities

7HVOD ZHE'DWD5HDGHU 76/$  \DKRR VWDUWHQG 

*HQHUDOB0RWRUV ZHE'DWD5HDGHU *0  \DKRR VWDUWHQG 
)RUG ZHE'DWD5HDGHU )  \DKRR VWDUWHQG 

• Plotting the volume (Fig. 7)

7HVOD> 9ROXPH @SORW ODEHO 7HVOD ILJVL]H  WLWOH 9ROXPH7UDGHG

*HQHUDOB0RWRUV> 9ROXPH @SORW ODEHO *0 
)RUG> 9ROXPH @SORW ODEHO ) 

B SOW[ODEHO 'DWH 

B SOW\ODEHO 9ROXPH 

SOWOHJHQG IDQF\ER[ 7UXHIUDPHDOSKD VKDGRZ )DOVHERUGHUSDG ORF XSSHUOHIW QFRO  

As a conclusion of the graph it can be observed that the volume of Ford

demonstrates that it has been traded considerably more than Tesla and
General Motors. The comparison could be clearer when compared to the
price (Fig. 8).
When comparing the closing price where Ford and General Motors
basically have behaved as stable, the volume of Ford has been exceed-
ingly variable. Given that Tesla is reasonably higher than Ford, the com-
parison of the present graph does not explain the reason between volume
and price (Fig. 9).
During the last five years, Ford´s security has had a bearish pat-
tern where the investors are believing less and less in the security. This
belief that the security will fall has caused a downtrend. The downtrend
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 131

Fig. 7 Volume traded of Tesla, General Motors and Ford (Source Elaborated by
the author with information from Yahoo Finance)

Fig. 8 Closing Price of Ford, General Motors and Tesla (Source Elaborated by
the author with information from Yahoo Finance)

explains the volume in which the investors could see an opportunity on

selling or buying (Fig. 10).
As seen in Fig. 21, the effect of the volume rises, and the effect is a
momentaneous rise in the stock price. This effect is important when ana-
lyzing trade because the volume confirms the trends (circle in red) but
132 M. GARITA

Fig. 9 Comparison of volume and closing price of Ford (Source Elaborated by

the author with information from Yahoo Finance)

has an important effect on the gains/losses of the investor. The impor-

tance of volume is that it can anticipate a price reversal, specifically when
with little movement in price there is a strong movement in volume. As
a part of technical analysis, it is important to understand this behavior
which can be combined with quantitative analysis that will be discussed
further in the book.
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 133

Fig. 10 Understanding volume and price with Ford Security (Source

Elaborated by the author with information from Yahoo Finance)

Candlestick Charts
The Candlestick charts were created in Japan during the 1700s by
Homma whose purpose was to analyze if there was a relation between
supply and demand of rice (Mitchell 2019). It is extremely useful for
analyzing emotional trading and actually is one of the most useful charts
in technical analysis.
134 M. GARITA

Fig. 11 Candlestick bar representation (Source Created by Bang [2019])

The Candlestick chart uses the open, high, low and close price in a
day. As stated in Fig. 22, the bar can be filled in or be black, although
the colors vary into green and red depending on the user. When the
body of the candlestick bar is filled, it means that the close was lower
than the open, therefore if the body is empty it means that the close was
higher than the open (Fig. 11).
To read a Candlestick chart it is important to understand if it is bullish
or bearish. These aspects are based on the price direction by analyzing
the close and open prices. It is assumed that the Candlesticks are respon-
sible for the focus on the opening price because of the importance of
these charts (J. J. Murphy 1999).
To create a Candlestick chart, the mplfinance1 package will be used.
The reason for using this package is the easiness of creating Candlestick
charts and how it adapts to the information accessed by Yahoo API.
The mplfinance package can be accessed at https://github.com/mat-
plotlib/mplfinance#release. There are different aspects considering this
package that is going to be highlighted during the process of creating
Candlestick Charts.

1 For more information concerning the mplfinance package visit: https://github.com/

matplotlib/mplfinance/blob/master/examples/customization_and_styles.ipynb.
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 135

Fig. 12 Zoom candlestick chart (Source Elaborated by the author with infor-
mation from Yahoo Finance)

• Importing libraries

import numpy as np
import pandas as pd
import pandas_datareader as pdr
import datetime
import mplfinance as mfp

• Setting the continuousdata range

start = datetime.datetime(2020,1,1)
end = datetime.datetime(2020,3,30)

• Choosing security

zoom = data.DataReader("ZM", 'yahoo', start, end)

• Plotting the candlestick (Fig. 12)

PISSORW ]RRPW\SH FDQGOH WLWOH =RRP&DQGOHVWLFNFKDUWWR 

\ODEHO &DQGOHVWLFN ILJUDWLR  ILJVFDOH  
136 M. GARITA

The above chart analyzes the trend concerning the security Zoom (ticker
ZM) which has seen an important growth given the COVID-19 and the
lockdown in various countries. From March 16 until March 30 the can-
dles have grown wider because of its high and low price, which demon-
strates high volatility in the security. A bullish pattern can also be seen
as well as the difference between closing price and opening price, which
leads to black and white candlesticks.

• Customizing the plot

– W\SH FDQGOH  determines the different choices that can be made to
create a chart. In the above example, FDQGOH was used but there are
other options such as OLQH RKOF  EDUV  RKFOBEDUV .
– title gives the possibility of creating a title in the graph to describe
what the candlestick or other graph is charting. It is important to
highlight that there are no year dates in the graphs, therefore cre-
ating a title with the expressed years are important.
– ylabel= &DQGOHVWLFN  sets a label in the y axis.
– figratio= () is for the author the ideal size for analysis, the
numbers can be altered to create a wider or narrower graph as
well as bigger or smaller.
– figscale=  is a conventional number for using in graphs.

Candlestick Charts and Volume

Candlesticks may also benefit from the analysis of volume. The purpose
of analyzing candlesticks and volume is based on the fact that the prices
are guided by the transactions, as exposed previously.

• Importing libraries

LPSRUWQXPS\DVQS
LPSRUWSDQGDVDVSG
LPSRUWSDQGDVBGDWDUHDGHUDVSGU
LPSRUWGDWHWLPH
LPSRUWPSOILQDQFHDVPIS

• Setting the continuous data range

VWDUW GDWHWLPHGDWHWLPH 
HQG GDWHWLPHGDWHWLPH 
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 137

Fig. 13 Dow Jones Candlestick chart with volume (Source Elaborated by the
author with information from Yahoo Finance)

• Choosing security

'RZB-RQHV GDWD'DWD5HDGHU A'-,$ \DKRR VWDUWHQG 

• Plotting the candlestick with volume (Fig. 13)

PISSORW 'RZB-RQHVW\SH FDQGOH WLWOH 'RZ-RQHV&DQGOHVWLFNFKDUWZLWKYROXPH

WR \ODEHO &DQGOHVWLFN ILJUDWLR  ILJVFDOH  YROXPH 7UXH  

As Fig. 24 demonstrates, there is a bearish pattern in the stock and

when analyzed with the volume, the shares traded have been growing
since February 14th, 2020, which is when the COVID-19 began to
have an impact in Spain and Italy. Also, on March 12 the World Health
Organization declared COVID-19 as a pandemic, which surged the vol-
ume and lowered considerably the price. Let us remember that according
to the Dow Jones Theory, the market discounts everything and this is
expressed in the price.
138 M. GARITA

Customizing Candlestick Charts and Volume

with **Kwargs

**Kwargs are extremely useful when working with functions and charts.
In the case of the mplfinance the kwargs can be useful to add the cus-
tomization of the plots into an only variable. Use kwargs in charts when
there are many variables rather than using an approach of describing each
variable, this is helpful for those reading the notebook (Mastromatteo
2020).

• Creating a **kwarg

For creating a **kwarg the process is simple because it is based on a dic-

tionary. To create a kwarg the process is as follows:

kwargs = dict(type='candle', title ='Zoom Candlestick chart -1/1/2020 to 30/3/2020',

ylabel='Candlestick',figratio=(30,15),figscale=0.75, volume=True) )

• Using the kwarg in a process (Fig. 14)

mfp.plot(Dow_Jones,**kwargs,style='yahoo')

The result of using kwargs is that it creates a dictionary, giving easy

access to the user when adapting the chart with the different formats
such as 'starandstripes', 'brasil' , 'mike' , 'charles' and 'classic'. The adaptation is user-
friendly and it allows for any alteration of the charts.

OHLC Charts with Volume

The OHCL charts follow the same idea behind the Candlestick charts.
Its name is derived from the Open price, High price, Closing price and
Lower price. An OHLC chart might be easy for some users to under-
stand when analyzing data, therefore it is included in the present book.
To interpret the OHLC chart, one must understand what each bar
means (Fig. 15).
When elaborating an OHLC chart using mplfinance it is a similar pro-
cess to the creation of the candlestick charts. Volume can also be added
as part of the chart and the customization of the chart is the same as the
candlestick.
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 139

Fig. 14 Candlestick chart and volume chart using colors (Source Elaborated by
the author with information from Yahoo Finance)

Fig. 15 OHLC bars explained (Source From the article written by Basurto
[2020])
140 M. GARITA

• Importing libraries

import numpy as np
import pandas as pd
import pandas_datareader as pdr
import datetime
import mplfinance as mfp

• Setting the continuous data range

VWDUW GDWHWLPHGDWHWLPH  

HQG GDWHWLPHGDWHWLPH  

• Choosing security

'RZB-RQHV GDWD'DWD5HDGHU A'-,$ \DKRR VWDUWHQG 

• Plotting the candlestick with volume (Fig. 16)

NZDUJV GLFW WLWOH 'RZ-RQHV2+/&FKDUWWR \ODEHO 2+/&

FKDUW ILJUDWLR  ILJVFDOH  YROXPH 7UXH  

PISSORW 'RZB-RQHV NZDUJV 

As show in Fig. 27 the process and the interpretation are similar to a

Candlestick chart. The volume has led to a change in prices and has had
an effect on the volatility of the security. The relation between shares
traded and the movement in price demonstrates the stated argument.

Line Charts with Volume

In one of the sections discussed earlier using matplotlib the analysis of
volume and a line chart was created by elaborating two charts. Using
mplfinace this can be done on the same chart.

• Importing libraries
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 141

Fig. 16 Dow Jones OHLC chart with volume (Source Elaborated by the
author with information from Yahoo Finance)

import numpy as np
import pandas as pd
import pandas_datareader as pdr
import datetime
import mplfinance as mfp

• Setting the continuous data range

start = datetime.datetime(2018,1,1)
end = datetime.datetime(2020,3,30)

• Choosing security

Dow_Jones = data.DataReader("^DJIA", 'yahoo', start, end)

142 M. GARITA

Fig. 17 Line charts with volume (Source Elaborated by the author with infor-
mation from Yahoo Finance)

• Plotting the candlestick with volume (Fig. 17)

kwargs = dict(title ='Dow Jones line chart with volume-1/1/2020 to 30/3/2020', ylabel='Line
chart',figratio=(30,15),figscale=0.75, volume=True) )

mfp.plot(Dow_Jones,**kwargs, type= 'line')

Moving Average with Matplotlib

The moving average is a technical indicator part of the technical analysis,
which means that it is centered on the trends of the stocks based on trad-
ing activity. It filters the noise concerning short-term prices and is useful
to identify a trend direction (Hayes 2020). The moving average (MA)
can be divided into simple moving average (SMA) or exponential mov-
ing average (EMA). Both of the indicators are useful for elaborating the
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 143

Moving Average Convergence Divergence (MACD) which is functional

for determining the momentum of a stock.
For creating an SMA, the companies that will be used are Amazon,
Walmart and Target. To calculate an SMA the first step is to determine
how many periods will be used as the average. The most common peri-
ods are 20, 50, 100 and 200 days (Milton 2020). This depends on the
data that is available and the purpose of the analysis. If there has been
a strong movement in the security market, it may be useful to use the
50-day period. If there has not been any change in the companies during
of comparison it may be useful to use the 100- or 200-day period.

• Installing packages

import numpy as np
import pandas as pd
from pandas_datareader import data as web
import pandas_datareader
import datetime
import matplotlib.pyplot as plt
%matplotlib inline

• Establishing dates of the analysis

start = datetime.datetime(2014, 1, 1)
end = datetime.datetime(2019, 1, 1)

• Choosing companies

Amazon = web.DataReader('AMZN', 'yahoo', start, end)

Walmart = web.DataReader('WMT', 'yahoo', start, end)
Target = web.DataReader('TGT', 'yahoo', start, end)

• Establishing the SMA for 50 days

Amazon['MA50'] = Amazon['Close'].rolling(50).mean()
Walmart['MA50'] = Walmart['Close'].rolling(50).mean()
Target['MA50'] = Target['Close'].rolling(50).mean()
144 M. GARITA

Amazon['MA100'] = Amazon['Close'].rolling(100).mean()
Walmart['MA100'] = Walmart['Close'].rolling(100).mean()
Target['MA100'] = Target['Close'].rolling(100).mean()

Amazon['MA200'] = Amazon['Close'].rolling(200).mean()
Walmart['MA200'] = Walmart['Close'].rolling(200).mean()
Target['MA200'] = Target['Close'].rolling(200).mean()

For the above example the rolling2 DataFrame from pandas was used.
By using rolling the window of days can be defined (in this case 50) and by
using the rolling with the mean, the result is the average of the 1st day to
the 50th day and then the average of 2nd day to the 51st day and so on.
Once the three companies have in the DataFrame the MA50 column,
then they can be plotted individually to analyze the momentum of the
stocks.

•Walmart Plot (Fig. 18)

Walmart['MA50'].plot(label='Walmart SMA50',figsize=(16,8))
Walmart['MA100'].plot(label='Walmart SMA100)
Walmart['MA200'].plot(label='Walmart SMA200)
Walmart['Close'].plot(label='Walmart Close')

_ = plt.xlabel('Date')

_ = plt.ylabel('Price')

_ = plt.title('Walmart SMA50 & Close Price Comparison')

plt.legend();

2 For more information concerning rolling visit: https://pandas.pydata.org/pan-

das-docs/stable/reference/api/pandas.DataFrame.rolling.html.
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 145

Fig. 18 Simple moving average of Walmart (Source Elaborated by the author

with information from Yahoo Finance)

• Target Plot (Fig. 19)

7DUJHW> 0$ @SORW ODEHO 7DUJHW60$ ILJVL]H 

7DUJHW> 0$ @SORW ODEHO 7DUJHW60$ 
7DUJHW> 0$ @SORW ODEHO 7DUJHW60$ 
7DUJHW> &ORVH @SORW ODEHO 7DUJHW&ORVH 

B SOW[ODEHO 'DWH 

B SOW\ODEHO 3ULFH 

B SOWWLWOH 7DUJHW60$ &ORVH3ULFH&RPSDULVRQ 

SOWOHJHQG 
146 M. GARITA

Fig. 19 Simple moving average of Target (Source Elaborated by the author

with information from Yahoo Finance)

• Amazon Plot (Fig. 20)

Amazon['MA50'].plot(label='Amazon SMA50',figsize=(16,8))
Amazon['MA100'].plot(label='Amazon SMA100)
Amazon['MA200'].plot(label='Amazon SMA200)
Amazon['Close'].plot(label='Amazon Close')

_ = plt.xlabel('Date')

_ = plt.ylabel('Price')

_ = plt.title('Amazon SMA50 & Close Price Comparison')

As seen on the plots, the momentum of Target and Amazon in the

selected dates, the SMA50 shows a bigger fall than the SMA 200. It is
important to notice that the three SMA predict a fall during the selected
dates. This could be seen as an indication that the security could fall. In
the case of Walmart, it is different since SMA50 and the SMA200 is that
the SMA200 demonstrates a stability in the stock.
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 147

Fig. 20 Simple moving average of Amazon (Source Elaborated by the author

with information from Yahoo Finance)

Moving Average with Mplfinance

Using the mplfinance for creating SMA is far more user-friendly than
matplotlib. The only inconvenience with the SMA in mplfinance is that at
the time this book is published, there are no options for adding legends.3
Other than the mentioned aspect, the process adds the characteristic
mav for moving average in which the number represents the different
periods.

• Importing libraries

import numpy as np
import pandas as pd
import pandas_datareader as pdr
import datetime
import mplfinance as mfp

3 The author contacted Daniel Goldfarb who is in charge of mplfinance and addressed

the issue in the following link: https://github.com/matplotlib/mplfinance/issues/21s.

148 M. GARITA

Fig. 21 Simple moving average for Walmart (Source Elaborated by the author
with information from Yahoo Finance)

• Setting the continuous data range

start = datetime.datetime(2014,1,1)
end = datetime.datetime(2019,1,1)

• Choosing security

Walmart = web.DataReader("^DJIA", 'yahoo', start, end)

• Plotting the candlestick with volume (Fig. 21)

kwargs = dict(title = 'Walmart SMA 50-100-200-1/1/4014 to 1/1/2010',

ylabel='SMA',figratio=(30,15),figscale=0.75, volume=True) )

mfp.plot(Walmart,**kwargs, type= 'line')

The SMA is equivalent to the one elaborated with matplotlib in Fig. 19:
Simple moving average of Walmart. The only problem are the legends,
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 149

but the process is far simpler. For a quick analysis it is recommended to

use the mplfinance as the main resource.

The Exponential Moving Average (EMA)

The exponential moving average (EMA) has the same background as the
moving average but it gives a specific weight to recent data points, which
is the most important difference. The EMA is also a technical indicator
that is useful to identify trends in the security market. The 50, 100 and
200 days are also used in the EMA as conventionalism (Hayes 2020).
When programming the EMA the Pandas DataFrame ewm for expo-
nential weighted is combined with the mean. This is an important aspect
since the EMA has an exponential weight that will determine the effect
on the curves and how it relates to the market. The process is done as
follows:

• Installing packages

import numpy as np
import pandas as pd
from pandas_datareader import data as web
import pandas_datareader
import datetime
import matplotlib.pyplot as plt
%matplotlib inline

• Establishing dates of the analysis

start = datetime.datetime(2014, 1, 1)
end = datetime.datetime(2019, 1, 1)

• Choosing companies

Amazon = web.DataReader('AMZN', 'yahoo', start, end)

Walmart = web.DataReader('WMT', 'yahoo', start, end)
Target = web.DataReader('TGT', 'yahoo', start, end)
150 M. GARITA

• Elaborating the EMA

Amazon['EMA50'] = Amazon['Close'].ewm(span=50, adjust=False).mean()

Amazon['EMA100'] = Amazon['Close'].ewm(span=100, adjust=False).mean()
Amazon['EMA200'] = Amazon['Close'].ewm(span=200, adjust=False).mean()

Walmart['EMA50'] = Walmart['Close'].ewm(span=50, adjust=False).mean()

Walmart['EMA100'] = Walmart['Close'].ewm(span=100, adjust=False).mean()
Walmart['EMA200'] = Walmart['Close'].ewm(span=200, adjust=False).mean()

Target['EMA50'] = Target['Close'].ewm(span=50, adjust=False).mean()

Target['EMA100'] = Target['Close'].ewm(span=100, adjust=False).mean()
Target['EMA200'] = Target['Close'].ewm(span=200, adjust=False).mean()

• Plotting the EMA of the three companies (Figs. 22, 23, and 24)

Amazon['EMA50'].plot(label='Amazon EMA50',figsize=(16,8))
Amazon['EMA100'].plot(label='Amazon EMA100')
Amazon['EMA200'].plot(label='Amazon EMA200')
Amazon['Close'].plot(label='Amazon Close')

_ = plt.xlabel('Date')

_ = plt.ylabel('Price')

_ = plt.title('Amazon EMA & Close Price Comparison')

plt.legend();
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 151

Fig. 22 Amazon EMA 50, 100, 200 (Source Elaborated by the author with
information from Yahoo Finance)

Fig. 23 Target EMA 50, 100,200 (Source Elaborated by the author with infor-
mation from Yahoo Finance)
152 M. GARITA

Fig. 24 Walmart EMA 50, 100,200 (Source Elaborated by the author with
information from Yahoo Finance)

Target['EMA50'].plot(label= Target EMA50',figsize=(16,8))

Target ['EMA100'].plot(label= Target EMA100')
Target ['EMA200'].plot(label= Target EMA200')
Target ['Close'].plot(label= Target Close')

_ = plt.xlabel('Date')

_ = plt.ylabel('Price')

_ = plt.title(Target EMA & Close Price Comparison')

plt.legend();
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 153

Walmart['EMA50'].plot(label= Walmart EMA50',figsize=(16,8))

Walmart ['EMA100'].plot(label= Walmart EMA100')
Walmart ['EMA200'].plot(label= Walmart EMA200')
Walmart ['Close'].plot(label= Walmart Close')

_ = plt.xlabel('Date')

_ = plt.ylabel('Price')

_ = plt.title(Walmart EMA & Close Price Comparison')

plt.legend();

When compared with the SMA the difference is that the fall of Target
and Amazon is not as drastic as in the SMA. This leads us to choose
between selecting the EMA strategy by analyzing the current effects on
the market or the SMA if the market has seen an important change in the
past few weeks and one thinks it should not have a weight in the analysis.

The Moving Average Convergence Divergence (MACD)

with Baseline

The Moving Average Convergence Divergence, better known as MACD,

was created by Gerard Appel with the purpose of understanding the
market behavior. The MACD is part of technical analysis and one of the
most used indicators when trading (Mitchell 2019).
The MACD is composed of three components4:

• The MACD line which measures the distance between two moving
averages.
• Signal line that identifies price change
• Histogram that represents the difference between MACD and sig-
nal line.

4 For more information visit the article: https://medium.com/duedex/what-is-macd-

4a43050e2ca8
154 M. GARITA

For calculating the MACD it is important to first calculate the

Exponential Moving Average (EMA). For this, it is useful to use the
Pandas DataFrame ewm that was used before. The equation to calculate
the MACD is the following:

Equation 2: MACD equation with EMA

MACD = EMA for 12 periods − EMA for 26 periods (2)
To input the equation in Python the ewm will beç combined with the
mean function to obtain the MACD. The span will be set in the first part
of the equation in 12 and the second part of the equation in 26. The
elaboration in Python is simple and is done as follows:

• Installing packages
LPSRUWQXPS\DVQS
LPSRUWSDQGDVDVSG
IURPSDQGDVBGDWDUHDGHULPSRUWGDWDDVZHE
LPSRUWSDQGDVBGDWDUHDGHU
LPSRUWGDWHWLPH
LPSRUWPDWSORWOLES\SORWDVSOW
PDWSORWOLELQOLQH

• Establishing dates of the analysis

VWDUW GDWHWLPHGDWHWLPH  

HQG GDWHWLPHGDWHWLPH  

• Choosing companies

$PD]RQ ZHE'DWD5HDGHU $0=1  \DKRR VWDUWHQG 

:DOPDUW ZHE'DWD5HDGHU :07  \DKRR VWDUWHQG 
7DUJHW ZHE'DWD5HDGHU 7*7  \DKRR VWDUWHQG 

• Creating the MACD

$PD]RQ> 0$&' @ $PD]RQ> &ORVH @HZP VSDQ DGMXVW )DOVH PHDQ 

$PD]RQ> &ORVH @HZP VSDQ DGMXVW )DOVH PHDQ 
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 155

:DOPDUW> 0$&' @ :DOPDUW> &ORVH @HZP VSDQ DGMXVW )DOVH PHDQ 

:DOPDUW> &ORVH @HZP VSDQ DGMXVW )DOVH PHDQ 

7DUJHW> 0$&' @ 7DUJHW> &ORVH @HZP VSDQ DGMXVW )DOVH PHDQ 

7DUJHW> &ORVH @HZP VSDQ DGMXVW )DOVH PHDQ 

• Creating the baseline

$PD]RQ> EDVHOLQH @ 
:DOPDUW> EDVHOLQH @ 
7DUJHW> EDVHOLQH @ 

• Plotting the MACD (Figs. 25, 26, and 27)

$PD]RQ> 0$&' @SORW ODEHO $PD]RQ0$&' ILJVL]H 

$PD]RQ> EDVHOLQH @SORW ODEHO %DVHOLQH 

B SOW[ODEHO 'DWH 

B SOW\ODEHO 3ULFH 

B SOWWLWOH $PD]RQ0$&' 

SOWOHJHQG 

Fig. 25 Amazon MACD with Baseline (Source Elaborated by the author with
information from Yahoo Finance)
156 M. GARITA

Fig. 26 Walmart MACD with Baseline (Source Elaborated by the author with
information from Yahoo Finance)

Fig. 27 Target MACD with baseline (Source Elaborated by the author with
information from Yahoo Finance)
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 157

Walmart['MACD'].plot(label='Walmart MACD',figsize=(16,8))
Walmart['baseline'].plot(label='Baseline')

_ = plt.xlabel('Date')

_ = plt.ylabel('Price')

_ = plt.title('Walmart MACD')

plt.legend();

7DUJHW> 0$&' @SORW ODEHO 7DUJHW0$&' ILJVL]H 

7DUJHW> EDVHOLQH @SORW ODEHO %DVHOLQH 

B SOW[ODEHO 'DWH 

B SOW\ODEHO 3ULFH 

B SOWWLWOH 7DUJHW0$&' 

SOWOHJHQG 

The interpretation of the MACD is set on the baseline which is zero.

When the MACD is above the baseline then the stock or market is bull-
ish, if the market is below the zero line then the market is bearish. The
MACD is also presented in a histogram but the line graph is a simpler
way to illustrate the process.

The Moving Average Convergence Divergence (MACD)

with Signal Line

Using a signal line in the MACD is one of the most important trading
items because of its interpretation.

– When the MACD crosses the signal line from below to above the
indicator is considered bullish.
– When the MACD crossed the signal line from above to below the
indicator is considered bearish (Posey 2019).
158 M. GARITA

The signal line is equivalent to an EMA of nine periods. The signal line
will be plotted with the MACD.

• Installing packages

import numpy as np
import pandas as pd
from pandas_datareader import data as web
import pandas_datareader
import datetime
import matplotlib.pyplot as plt
%matplotlib inline

• Establishing dates of the analysis

start = datetime.datetime(2014, 1, 1)
end = datetime.datetime(2019, 1, 1)

• Choosing companies

Target = web.DataReader('TGT', 'yahoo', start, end)

• Creating the EMA for 12 and 26 periods

ema_12 = Target['Close'].ewm(span=12, adjust=False).mean()

ema_16 = Target['Close'].ewm(span=26, adjust=False).mean()

• MACD and Signal Line (9 periods)

macd = ema_12 - ema_16

signal_line = macd.ewm(span=9, adjust=False).mean()

• Plotting the MACD and the Signal Line (Fig. 28)

 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 159

Fig. 28 MACD and Signal Line (Source Elaborated by the author with infor-
mation from Yahoo Finance)

macd.plot(label='Target MACD',figsize=(16,8))
signal_line.plot(label='Baseline')

_ = plt.xlabel('Date')

_ = plt.ylabel('MACD and Signal Line')

_ = plt.title('Target MACD and Signal Line')

plt.legend();
160 M. GARITA

Bollinger Bands ®
When John Bollinger created the Bollinger Bands in the 1980s he found
a means of joining the quantitative aspect (standard deviation) and tech-
nical analysis for decision-making. Bollinger bands combine standard
deviation, a measure for volatility, and moving average defining when the
security has a contraction or an expansion (Bollinger 2018).
Bollinger Bands are extremely useful when analyzing security because:

– When there is low volatility the bands will be close together. When
there is high volatility the bands will be apart. Periods of low vola-
tility are often followed by high volatility. The same is for periods of
high volatility followed by low volatility.
– If a price moves beyond the upper barrier the prices are considered
overbought. Meaning that the stock is being bought at unjustifiably
high prices.
– If a price moves below the upper barrier then the prices are con-
sidered oversold. Meaning that the stocks are selling below its true
value.

To calculate a Bollinger Band there is a need for a window of days for

the average and a number of standard deviations for the higher and
lower bands. The number of days conventionally is set to 20 days. For
the Bollinger Band the package that will be used is the mplfinance.

• Importing libraries

LPSRUWQXPS\DVQS
LPSRUWSDQGDVDVSG
LPSRUWSDQGDVBGDWDUHDGHUDVSGU
LPSRUWGDWHWLPH
LPSRUWPSOILQDQFHDVPIS

• Setting the continuous data range

VWDUW GDWHWLPHGDWHWLPH 

HQG GDWHWLPHGDWHWLPH 
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 161

• Choosing security

'RZB-RQHV GDWD'DWD5HDGHU A'-,$ \DKRR VWDUWHQG 

• Creating the Bollinger Bands

ZLQGRZBRIBGD\V 
QXPEHUBVWG 

rolling_mean = Dow_Jones['Close'].rolling(window_of_days).mean()
rolling_std = Dow_Jones['Close'].rolling(window_of_days).std()

Dow_Jones['Rolling Mean'] = rolling_mean

Dow_Jones['Bollinger High'] = rolling_mean + (rolling_std * number_std)
Dow_Jones['Bollinger Low'] = rolling_mean - (rolling_std * number_std)

• Plotting the Bollinger Bands (Fig. 29)

high_low = Dow_Jones[['Bollinger Low','Bollinger High']]

apd = mfp.make_addplot(high_low)

kwargs = dict(title ='Dow Jones Bollinger Band 1/8/2019 to 1/4/2020', ylabel='Bollinger Bands',
figratio=(30,15),figscale= 0.75)

mfp.plot(Dow_Jones, addplot = apd, **kwargs, type = 'line')

As expressed before, there are three breaking points in the lower band
during the COVID-19 pandemic crisis, which means that the stocks
are oversold. This leads to the argument that because of fear there was
dumping or selling securities before they lost value.
162 M. GARITA

Fig. 29 Dow Jones Bollinger Bands (Source Elaborated by the author with
information from Yahoo Finance)

Backtesting Strategies for Trading

Backtesting has become extremely useful with Python and API data
because it allows historical information to be accessed easily, as seen
before, and be used nn a prompt manner. Backtesting for trading is
important because it offers a strategy for creating returns based on the
performance of a stock.

Parabolic SAR
The parabolic SAR gives and edges to the traders given that it analyzes
the movement of the stock. It was created by J. Welles Wilder Jr., which
also created the RSI (C. Murphy 2020). The logic behind the parabolic
SAR is as follows:

Equation 3: Uptrend and Downtrend SAR Equation

Uptrend Parabolic SAR = Prior SAR
+ Prior Acceleration Factor ∗ (Prior Extreme Point − Prior SAR)
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 163

Downtrend Parabolic SAR = Prior SAR − Prior Acceleration Factor ∗

(Prior SAR − Prior Extreme Point)
(3)
There are important aspects when creating a parabolic SAR, for exam-
ple, the extreme point in the uptrend is the highest price, usually refer as
High, and in the Downtrend, it is the low price referred to as Low. The
acceleration factor is usually set at 0.02. The acceleration factor affects
the SAR when the extreme point is recorded by 0.02. The acceleration
factor is modified depending on each trader and its strategy.
When using the TA-Lib library for calculating the parabolic SAR the
function SAR will be used to retrieve the information. It will ask for the
following:

– Choose the High price for the extreme points in uptrend

– Choose the Low price for the extreme point in downtrend.
– Choose the acceleration factor (as recommended 0.02)
– Choose the maximum for the acceleration factor (recommended 0.02)

The implementation is as follows:

– Import packages including talib

import talib
import numpy as np
import pandas as pd
import pandas_datareader as pdr
from pandas_datareader import data as web
import datetime
import mplfinance as mpf
import matplotlib.pyplot as plt
%matplotlib inline

– Choose dates

start = datetime.datetime(2019,8,1)
end = datetime.datetime(2021,1,30)

– Choose a stock

aapl = web.DataReader("aapl", 'yahoo', start,end)

164 M. GARITA

Fig. 30 AAPL Parabolic SAR

– Configure the parabolic SAR

aapl['SAR'] = talib.SAR(aapl['High'].values, aapl['Low'].values,

acceleration=0.02, maximum=0.2)
aapl

Date High Low Open Close Volume Adj Close SAR

2019-08- 54.507500 51.685001 53.474998 52.107498 216,071,600.0 51.311756 NaN
01
2019-08- 51.607498 50.407501 51.382500 51.005001 163,448,400.0 50.226093 54.507500
02
2019-08- 49.662498 48.145000 49.497501 48.334999 209,572,000.0 47.596859 54.425500
05
2019-08- 49.517502 48.509998 49.077499 49.250000 143,299,200.0 48.497894 54.174280
06
2019-08- 49.889999 48.455002 48.852501 49.759998 133,457,600.0 49.000103 53.933109
07

– Plot the Parabolic SAR with the closing price (Fig. 30).

aapl['Close'].plot(label='Closing Price',figsize=(16,8))
aapl['SAR'].plot(label= 'SAR')

_ = plt.xlabel('Date')

_ = plt.ylabel('Parabolic SAR and Closing Price')

_ = plt.title('AAPL Parabolic SAR')

plt.legend();
 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 165

Table 1 Closing time in stock markets

Stock exchange Closing time Time zone

New York Stock Exchange (NYSE) 4:00 p.m Eastern Time

London Stock Exchange (LSE) 4:30 p.m–4:35 p.m Greenwich Mean Time
Bolsa de Madrid 5:30 p.m–5:45 p.m Greenwich Mean Time + 1

Source Created by the author with information from (Bolsa de Madrid 2020) (London Stock Exchange
Group 2020) (NYSE 2020)

The parabolic SAR is useful to buy or sell stocks depending on its rela-
tionship with the closing price of the stock, in this example, it is the
stock of Apple. The parabolic SAR recommends us to buy if the SAR line
is below the closing price and to sell if the SAR is above the closing price.
As seen before, at the beginning of the chart, the SAR line and the clos-
ing price were very similar but as volatility has struck Apple, the oppor-
tunity for selling and buying is clearer. This also can be complemented
with the stochastic oscillator.

Fast and Slow Stochastic Oscillators

Stochastic oscillators are momentum indicators, such as the SMA seen
before or the SAR. The difference considering the stochastic oscillators
is that it can be divided into fast and slow, and that it is considered usu-
ally during a period of fourteen days. The formula for calculating the fast
stochastic oscillator, referred to as %K, is as follows:

Equation 4: Fast Stochastic Oscillator Equation

100 ∗ Closing price − Low price of the 14 previous trading sessions
%K = .
(Highest price in the last 14 day sessions − Low price of the 14 previous trading sessions
(4)
The interpretation regarding %K is that if the result is 80 then the
price is 8% above the prices in the last 14 days. The days can change
based on the intuition and knowledge of the trader and it is usual to see
a five-day period.
166 M. GARITA

The slow indicator creates a change by applying a three(3)-day mov-

ing average to fast calculation (%K). The result is that the slow stochastic
oscillator (%D) creates a signal line that is useful to know when to buy
and when to sell. The process for callating a fast oscillator is as follow:

– Change the dates

start = datetime.datetime(2020,12,1)
end = datetime.datetime(2021,1,30)

It is important to create an oscillator with fewer data points than the

one’s used before. This is extremely useful for noticing the effect.

– Fast stochastic oscillator

aapl['fastk'], aapl['fastd'] = talib.STOCHF(aapl['High'].values,

aapl['Low'].values, aapl['Close'].values, fastk_period=14, fastd_period=3)

In the equation before, the standard fourteen days were used for the %K
and three days for the %D.

– Slow stochastic oscillator

aapl['slowk'], aapl['slowd'] = talib.STOCH(aapl['High'].values,

aapl['Low'].values, aapl['Close'].values, fastk_period=14,
slowk_period=3, slowd_period=3)

– Result
Date High Low Open Close Volume Adj Close SAR slowk slowd fastk fastd
2021-01-25 145.089996 136.539993 143.070007 142.919998 157,611,700 142.919998 127.173966 87.180935 71.431681 88.401933 87.180935
2021-01-26 144.300003 141.369995 143.600006 143.160004 98,390,600 143.160004 128.248927 90.765333 85.085841 89.684699 90.765333
2021-01-27 144.300003 140.410004 143.429993 142.059998 140,843,800 142.059998 129.259392 87.155227 88.367165 83.379048 87.155227
2021-01-28 141.990005 136.699997 139.520004 137.089996 142,621,100 137.089996 130.209228 76.393343 84.771301 56.116282 76.393343
2021-01-29 136.740005 130.210007 135.830002 131.960007 177,180,600 131.960007 145.089996 55.823745 73.124105 27.975904 55.823745
ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON
167
168 M. GARITA

– Plotting results (Figs. 31 and 32)

aapl['slowk'].plot(label='slowk',figsize=(16,8))
aapl['slowd'].plot(label= 'slowd')

_ = plt.xlabel('Date')

_ = plt.ylabel('Slow Stochastic Oscillator')

_ = plt.title('Slow Stochastic Oscillator - AAPL')

plt.legend();

aapl['fastk'].plot(label='fastk',figsize=(16,8))

aapl['fastd'].plot(label= 'fastd')

_ = plt.xlabel('Date')

_ = plt.ylabel('Fast Stochastic Oscillator')

_ = plt.title('Fast Stochastic Oscillator - AAPL')

plt.legend();

Fig. 31 Slow stochastic oscillator

 ELEMENTS FOR TECHNICAL ANALYSIS USING PYTHON 169

Fig. 32 Fast stochastic oscillator

References
Bang, Julie. 2019. Candlestick bar. Investopedia.
Basurto, Stefano. 2020. Python trading toolbox: Introducing OHLC charts,
7 January. Accessed March 31, 2020. https://towardsdatascience.com/
trading-toolbox-03-ohlc-charts-95b48bb9d748.
Bollinger, John. 2018. John Bollinger answers “What are Bollinger Bands?”, n.d.
Accessed March 2, 2020. https://www.bollingerbands.com/bollinger-bands.
Bolsa de Madrid. 2020. Electronic Spanish Stock Market Interconnection System
(SIBE), n.d. Accessed March 23, 2020. http://www.bolsamadrid.es/ing/
Inversores/Agenda/HorarioMercado.aspx.
Chen, James. 2019. Line Chart, 12 August. Accessed March 25, 2020. https://
www.investopedia.com/terms/l/linechart.asp.
Halton, Clay. 2019. Line Graph, 21 August. Accessed March 25, 2020. https://
www.investopedia.com/terms/l/line-graph.asp.
Hayes, Adam. 2018. Volume definition, 4 February. Accessed March 25, 2020.
https://www.investopedia.com/terms/v/volume.asp.
Hayes, Adam. 2020a. Exponential Moving Average - EMA definition, 8 July.
Accessed April 2, 2020. https://www.investopedia.com/terms/e/ema.asp.
Hayes, Adam. 2020b. Moving Average (MA), 31 March. Accessed March 31,
2020. https://www.investopedia.com/terms/m/movingaverage.asp.
Kevin, S. 2015. Security analysis and portfolio management. Delhi: PHI.
London Stock Exchange Group. 2020. London Stock Exchange Group Business
Day, n.d. Accessed March 25, 2020. https://www.lseg.com/areas-expertise/
170 M. GARITA

our-markets/london-stock-exchange/equities-markets/trading-services/
business-days.
Mastromatteo, Davide. 2020. Python args and kwargs: Demystified. 09 September.
Accessed March 31, 2020. https://realpython.com/python-kwargs-and-args/.
Milton, Adam. 2020. Simple, exponential, and weighted moving averages, 9
November. Accessed March 31, 2020. https://www.thebalance.com/
simple-exponential-and-weighted-moving-averages-1031196.
Mitchell, Cory. 2019. Understanding basic candlestick charts. 19 December.
Accessed March 30, 2020. https://www.investopedia.com/trading/
candlestick-charting-what-is-it/.
Mitchell, Cory. 2020. How to use volume to improve your trading, 25 February.
Accessed March 27, 2020. https://www.investopedia.com/articles/techni-
cal/02/010702.asp.
Murphy, Casey. 2020. Investopedia, 16 November. Accessed January 1, 2021.
https://www.investopedia.com/trading/introduction-to-parabolic-sar/.
Murphy, John J. 1999. Technical analysis of the financial markets. New York:
New York Institute of Finance.
NYSE. 2020. TAQ closing prices, n.d. Accessed March 25, 2020. https://www.
nyse.com/market-data/historical/taq-nyse-closing-prices.
Posey, Luke. 2019. Implementing MACD, 30 March. Accessed April 2, 2020.
https://towardsdatascience.com/implementing-macd-in-python-cc9b2280126a.
 Valuation and Risk Models with Stocks

Abstract One of the most important aspects for the analysis of securi-
ties is to analyze its risk models and how to valuate the instrument. In
the present chapter aims to explain the importance of risk, the different
financial measures for understanding risk in a portfolio of securities and
the impact on the returns of a portfolio.

Keywords Beta · Alpha · Risk · Valuation · Portfolio

This part of the book is centered on the management of risk. In this

aspect, financial risk should be seen as the risk of the financial markets
based on liquidity, operational and strategical risk. As seen before there
are three tools for managing risk being fundamental analysis, technical
analysis and quantitative analysis (Chen 2019).
Therefore, this section of the book is centered on Modern Portfolio
Theory, Value at Risk and Monte Carlo Simulations. These three methods
are imperative for understanding the financial markets and risk management.

Creating a Portfolio
Managing portfolios with multiple assets is one of the most interesting
processes when working on finance. Given that there are different assets
which can compose a portfolio, one of the most important processes of

© The Author(s), under exclusive license to Springer 171

Nature Switzerland AG 2021
M. Garita, Applied Quantitative Finance,
https://doi.org/10.1007/978-3-030-29141-9_7
172 M. GARITA

establishing a portfolio is identifying its risk. For understanding how to

build and analyze a portfolio this chapter will center on the process for
creating, optimizing and evaluating a portfolio using Python.
When creating the portfolio there are different approaches that can
be taken. There is always the portfolio that is created by the sentiment of
the investor, the companies that they believe are the correct ones and the
risk that he is willing to take. Considering risk, it depends if the inves-
tor is interested in bonds, cash, stocks or equivalents. In the case of this
example, the most popular approach is to determine the market value
(Bryant 2020).
The first step is to determine the companies that are performing in the
index or choosing the companies that are of interest to the investor. For
example, Disney, Netflix, Tesla and Amazon have been chosen. To the
present time that the book was written the companies have been behav-
ing greatly and creating strong returns.

• Creating a Portfolio with 4 variables

start = datetime.datetime()
end = datetime.datetime()

tickers = ( 1)/;  ',6  76/$  $0=1 )

stocks = pd.DataFrame()
IRU[ LQ WLFNHUV
     stocks[x] = web.DataReader(x, ‘yahoo’, start, end)[‘Close’]

The second step then creating the portfolio is assigning the weights that
the portfolio must have. For this approach, considering market capitali-
zation is important.

• Determining market value

To determine market value two steps are essential. To know the last price
of the security traded and the number of shares outstanding. For this, it
is important to have a benchmark that can be used as a reference for the
securities. Since the example establishes the last price as of December 26,
2018, then the prices can be obtained as follows:

stocks.tail()
 VALUATION AND RISK MODELS WITH STOCKS 173

Table 1 Shares outstanding of Tesla, Netflix, Amazon and Walt Disney

Company Shares Oustanding1

Tesla Inc 172,721,000

Netflix Inc 436,599,000
Amazon.com, Inc 491,203,000
Walt Disney Company (The) 1,490,777,000

Source Elaborated by the author with information from Yahoo Finance and Nasdaq aObtained from the
https://www.nasdaq.com

Table 2 Market capitalization of Netflix, Tesla, Amazon and Walt Disney

Company Last Stock Price Shares Oustanding Market Capitalization

Telsa Inc 326.09 172,721,000 56,322,590,890

Netflix Inc 253.67 436,599,000 110,752,068,330
Amazon.com, Inc 1470.90 491,203,000 722,510,492,700
Walt Disney Company 105.83 1,490,777,000 157,768,929,910
(The)

Source Elaborated by the author with information from Yahoo Finance and Nasdaq

Date NFLX DIS TSLA AMZN

2018–12-26 253.669998 105.830002 326.089996 1470.900024
2018–12-27 255.570007 106.519997 316.130005 1461.640015
2018–12-28 256.079987 107.300003 333.869995 1478.020020

According to the NASDAQ, the shares outstanding of the company

are as follows (Table 1):
To calculate the market capitalization of each company, the process is
to multiply the last stock price with the shares outstanding (Table 2).
With the information, the next process is to obtain the market capitaliza-
tion of the benchmark, in this case the Standard and Poor’s 500 (S&P 500).
The market capitalization for December 31, 2018 is 21.03 trillion (Table 3).
With the information above it is easier to understand the market value
of the stock and the relation between the benchmark and the stock that
has been chosen for the portfolio. In this case, the most significant stock
in the portfolio is Amazon with a 69.98%.
For example of the portfolio, it may be important to add other com-
panies that can be included to weigh down the participation of Amazon,
but for this example the assigning of weights is as follows:

1 Obtained from the https://www.nasdaq.com

174 M. GARITA

Table 3 Market capitalization and portfolio weight

Company Market Capitalization Portfolio Weight of the

S&P 500

Telsa Inc 56,322,590,890 5.38%

Netflix Inc 110,752,068,330 10.57%
Amazon.com, Inc 722,510,492,700 68.98%
Walt Disney Company (The) 157,768,929,910 15.06%
Total 1,047,354,081,830 100%

Source Elaborated by the author with information from Yahoo Finance and Nasdaq

• Calculate stocks returns

stocks_return = (stocks/stocks.shift()

• Calculate the weights of the portfolio based on market capitalization

portfolio_weights = np.array([])

• Create a weighted return portfolio

weighted_returns_portfolio = stocks_return.mul(portfolio_weights, axis = )

weighted_returns_portfolio.head().dropna()

• Create a stock portfolio based on the weights assigned

stocks_return['Portfolio' ] = weighted_returns_portfolio.sum(axis = 1 ).dropna()

Calculating Statistical Measures on a Portfolio

Now that the portfolio is created there are certain statistical measures
that can be calculated. The first one is the portfolio standard deviation
which is as follows:

Equation 1: Portfolio standard deviation


σPortfolio = wt ∗ � ∗ w

wt= transposed portfolio of the weights assigned

 = the covariance matrix of the returns
w = weights
 VALUATION AND RISK MODELS WITH STOCKS 175

The first step is to obtain the annual covariance. For the annual covariance
the number of days that will be used is 252, a more conventional approach to
the 360 or 365. The days are based on the stock market available days:

• Calculate the covariance of the stocks and the portfolio

stocks_covariance = stocks_return.cov()

Covariance Matrix of the Stock and the Portfolio

NFLX DIS TSLA AMZN Portfolio

NFLX 0.000735 0.000089 0.000247 0.000236 0.000255
DIS 0.000089 0.000142 0.000090 0.000080 0.000094
TSLA 0.000247 0.000090 0.000797 0.000186 0.000601
AMZN 0.000236 0.000080 0.000186 0.000381 0.000207
Portfolio 0.000255 0.000094 0.000601 0.000207 0.001265

• Annualize the covariance

annualized_covariance = stocks_covariance * 252

annualized_covariance

Annualized Covariance of the Stocks and the Portfolio

NFLX DIS TSLA AMZN Portfolio

NFLX 0.185274 0.022497 0.062368 0.059461 0.064322
DIS 0.022497 0.035697 0.022763 0.020037 0.023703
TSLA 0.062368 0.022763 0.200928 0.046951 0.151433
AMZN 0.059461 0.020037 0.046951 0.096091 0.052175
Portfolio 0.064322 0.023703 0.151433 0.052175 0.318702

• Obtain the standard deviation of the portfolio

portfolio_sd = np.sqrt(np.dot(portfolio_weights.T, np.dot2(annualized_

covariance, portfolio_weights)))
portfolio_sd

0.34392103687572534

2 The np.dot converts the two data sets into one. For more information visit: https://

docs.scipy.org/doc/numpy/reference/generated/numpy.dot.html.
176 M. GARITA

The volatility of the portfolio is 34.49% with can be considered as a

highly volatile portfolio. This is important because when analyzing the
beta, the behavior of the market and the stock will be included into the
analysis.

The Capital Asset Pricing Model

The Capital Asset Pricing Model better known as CAPM is a model
created by William Sharpe based on the work of Harry Markowitz. Its
major assumptions are that the offer of financial assets is equal to the
demand of financial assets (Mullins 1982). When the assumption is ana-
lyzed under perfect competitiveness, which means that the price is deter-
mined by the offer and demand of economic agents, hence determine
the price of the asset.
Another important aspect of the CAPM is that it only considers the
systematic risk (market risk). Systematic risk can be understood as the
risk that is not diversifiable and therefore cannot be reduced (Fontinelle
2019). Risk, as a whole is divided as follows:

Equation 2: Total risk

Total Risk = systematic risk + non systematic risk

The Beta
The measurement of the systematic risk is through the beta, which is a
degree of sensitivity that includes the variation of an asset compared with
an index that is used as a benchmark.

Equation 3: Beta calculation with correlation

σi
βi = ρim
σm
ρim = Correlation between asset i and the market
σi = variance of asset i
σm = variance of the market

Another way of calculating the beta is through the following formula:

 VALUATION AND RISK MODELS WITH STOCKS 177

Equation 4: Beta calculation with covariance

cov(i,m)
βi =
σm
cov(i,m)= covariance of assets i and market
σm= variance of the market

Given that the covariance will be obtained for calculating the beta, it
is important to elaborate a covariance matrix. First, the process will be
explained with the covariance formula and then it will be integrated with
the correlation formula.
Before coding in Python there are certain aspects that should be clar-
ified to understand the process of computing the beta with covariance:

• Using a for loop: The program below uses a simple for loop because
there is a need here to create a DataFrame with the two variables.
The process is simple, and for more information there are different
ways a for loop3 can be used. See For loop.
• Choosing a market for comparison: For example, the S&P 500
(^GSPC) was used because it is an interesting market reference
when trying to determine the behavior of the company.
• Using iloc: iloc4 is a pandas DataFrame function that selects a posi-
tion. It can be used to slice the DataFrame or to create new col-
umns. Please visit the chapter titled The Basics for more information.
• Using 252 days: The concept of 252 days comes from the regular
trading hours (RTH). It is a rule of thumb (Mitra y Mitra 2011).
The covariance and the variance are multiplied by this to convert it
into daily.

These aspects are important since they will be repeated throughout the
following sections of the book.

3 An excellent resource for understanding for loops can be found here: https://www.

w3schools.com/python/python_for_loops.asp.
4 For more information on using iloc and examples visit: https://pandas.pydata.org/

pandas-docs/stable/reference/api/pandas.DataFrame.iloc.html.
178 M. GARITA

• Installing packages

import numpy as np

import pandas as pd

from pandas_datareader import data as web

import pandas_datareader

import datetime

import matplotlib.pyplot as plt

% matplotlib inline

• Setting time and date

• start = datetime.datetime(2014, 1, 1 )
• end = datetime.datetime(2019, 1, 1 2019, 1, 1)

• Using a for loop for obtaining tickers

tickers = ['DIS', '^GSPC' ]
stocks = pd.DataFrame()
for x in tickers:
stocks[x] = web.DataReader(x, 'yahoo' , start, end)['Close' ]

• Creating logarithmic returns

stocks_return = np.log(stocks/stocks.shift(1 ))

• Calculating the covariance matrix

covariance = stocks_return.cov() * 252

covariance

Covariance Matrix

DIS ^GSPS
DIS 0.035899 0.015926
^GSPS 0.015926 0.017540
 VALUATION AND RISK MODELS WITH STOCKS 179

• Setting the market covariance into a variable

covariance_market = covariance.iloc[0,1 ]
covariance_market

• Calculating the variance of the market

market_variance = stocks_return[' ^GSPC' ].var() * 252

market_variance

• Calculating Beta with covariance

beta_Disney = covariance_market/market_variance
beta_Disney

Beta = 0.90796359355848766

For calculating the beta with correlation, the formula, it is easier but it
is less common. It is important to remember that when comparing beta,
the same days of data are being used, and that this is applied to all the
variables. There are different approaches to the beta but the most nota-
ble aspect is to understand the result. To.

• Installing packages

import numpy as np

import pandas as pd

from pandas_datareader import data as web

import pandas_datareader

import datetime

import matplotlib.pyplot as plt

% matplotlib inline
180 M. GARITA

• Setting time and date

start = datetime.datetime(2014, 1, 1 )
end = datetime.datetime(2019, 1, 1 )

• Using a for loop for obtaining tickers

tickers = [NFLX, '^GSPC' ]
stocks = pd.DataFrame()
for x in tickers:
    stocks[x] = web.DataReader(x, 'yahoo' , start, end)[ 'Close' ]

• Creating logarithmic returns

stocks_return = np.log(stocks/stocks.shift(1 ))

• Calculating correlation

correlation = stocks_return.corr()
correlation

Correlation Matrix
DIS NFLX
NFLX 1.000000 0.467229
^GSPS 0.467229 1.000000

• Inserting Correlation into a variable

correlation_NFLX_GSPC = correlation.iloc[0,1 ]
correlation_NFLX_GSPC

• Netflix variance

variance_NFLX = stocks_return['NFLX' ].var()

variance_NFLX

• Market variance

variance_GSPC = stocks_return['^GSPC' ].var()

variance_GSPC
 VALUATION AND RISK MODELS WITH STOCKS 181

• Beta

beta = correlation_NFLX_GSPC * (variance_NFLX/variance_NFLX)

beta

Beta = 0.46722863704340151

The beta above is annualized since the correlation is obtained by

using all the data and comparing it. This is why the first formula is
used more often because it is capable of representing the beta using
daily information. Notice that correlation can change if the data is used
daily, monthly or annually and this could lead to results that are not
comparable.
The next step is to interpret the beta coefficient. The interpretation of
the beta coefficient is very useful since it helps understand the relation
between a security and the market it trades. This is fundamental when
analyzing a portfolio because it is imperative to comprehend what hap-
pens when the market moves with a certain security that is included in
the portfolio (Table 4).
The beta coefficients results:

Beta Disney = 0.9079

Beta Netflix = 0.4672

Both of them are less volatile than the market, meaning that if the
S&P 500 shifts in an upward direction by 1% then Disney will shift by
0.91% and Netflix by 0.46%. The beta is important for analyzing the pro-
cess in which the assets behave when compared to a market.

Table 4 The Beta Table

βi = 1 The asset is exactly as volatile as the market it is compared with

βi >1 The asset is more volatile than the market it is compared with
βi <1>0 The asset is less volatile than the market it is compared with
βi =0 The asset is not correlated with the market it is compared with
βi <0 Negatively correlated with the market it is compared with
182 M. GARITA

The Beta and the CAPM

The CAPM is considered part of the factor analysis because it allows the
understanding of the relationship between variables. It is important to
consider that when using the factor analysis, the key aspect understand-
ing the relation of the correlated variables with the factors.
The first step for developing the CAPM is calculate the excess returns.
An excess return happens when the asset or portfolio exceeds the risk-
free return. The equation is as follows:

Equation 5: Excess return

Excess Return = Return − Risk Free Return
A risk-free return can be defined as an investment that with zero risk
(lowest risk possible) guarantees a return. Usually for the risk-free rate
the U.S. Treasury Bills (known as T-Bills) are considered because they
are backed by the U.S. Government and the risk of default is minimum.
To understand the relation between the Risk-Free Return, the
Beta and the CAPM the easiest way is to understand the model by its
equation:

Equation 6: Capital Asset Pricing Model

 
E Rp − RF = βp (E(Rm ) − RF)

E Rp − RF = The excess in the expected return of Portfolio P

 

E(Rm ) − RF = The excess expected return of market portfolio

RF = Risk-Free Return
βp = Beta of the portfolio

According to the equation, the first step for calculating the CAPM is
to create a portfolio. To create a portfolio, one must understand that it is
composed of different assets and that those assets should have a different
weight.
The weights are known as portfolio weights and there are different
methods for calculating the weights but the example will focus on the
market value method seen before.
 VALUATION AND RISK MODELS WITH STOCKS 183

• Installing packages

import numpy as np

import pandas as pd

from pandas_datareader import data as web

import pandas_datareader

import datetime

import matplotlib.pyplot as plt

% matplotlib inline

• Creating a Portfolio with 4 variables

start = datetime.datetime(2014, 1, 1 )
end = datetime.datetime(2019, 1, 1 )
tickers = ['NFLX', 'DIS', 'TSLA', 'AMZN' ]
stocks = pd.DataFrame()for x in tickers:
stocks[x] = web.DataReader(x, ‘yahoo’, start, end)[‘Close’]

The second step when building a portfolio is identifying the weights

that the model will consider. For this it is important to know how much
money is going to be invested in the model. For this example, the invest-
ment will be of USD 500,000 and it will be distributed according to the
following information (Table 5):
Considering the above, the subsequent step is to be familiar with the
construction of the portfolio based on the company that is going to be
chosen. One of the key aspects is that the sum of the portfolio weights

Table 5 Investing USD 500,000

Company Investment Portfolio Weight

Telsa Inc 130,000 26%

Netflix Inc 100,000 20%
Amazon.com, Inc 170,000 34%
Walt Disney Company (The) 100,000 20%
Total 500,000 100%
184 M. GARITA

has to add to a 100% or to 1 depending on if you are using decimal or

percentage. This is important because of the portfolio return formula
given below:

Equation 7: Portfolio Return

Rp = Ra1 wa1 + Ra2 wa2 + Ra3 wa3 + . . . + Ran wax
Rp = Return of the portfolio
Ra1 = Return of the first asset
wa1 = Weight of the first asset.

• Weights for the portfolio

portfolio_weights = np.array([0.30, 0.20, 0.25, 0.25 ])

• Calculating return for each company on the portfolio

portfolio_weights = np.array([0.30, 0.20, 0.25, 0.25 ])

• Creating a portfolio with four companies using close price for a

time series between January 1, 2014 to January 1, 2019

start = datetime.datetime(2014, 1, 1 )
end = datetime.datetime(2019, 1, 1 )

tickers = ['NFLX', 'DIS', 'TSLA', 'AMZN' ]

stocks = pd.DataFrame()
for x in tickers:
stocks[x] = web.DataReader(x, 'yahoo' , start, end)[ 'Close' ]

• Create logarithmic returns in the portfolio

stocks_return = np.log(stocks/stocks.shift(1 ))

• Drop missing values in the series and visualize the data

stocks_return.dropna().head()
 VALUATION AND RISK MODELS WITH STOCKS 185

• Calculate the weighted stock returns of the portfolio

weighted_returns_portfolio = stocks_return.mul(portfolio_weights,
axis = 1 )

• Calculating the returns of a portfolio

stocks_return['Portfolio' ] = weighted_returns_portfolio.sum(axis = 1 )

• Calculate the cumulative returns of the portfolio and plot them

c u m u l a t i v e _ r e t u r n s _ p o r f o l i o = ( ( 1 + s t o c k s _ r e t u r n [ 'Portfolio' ] ) .
cumprod()-1).

cumulative_returns_porfolio.plot(label = 'Cumulative Returns of the Portfolio' ,fig-

size = (19,8),title = ‘Cumulative Returns’)
_ = plt.xlabel('Date' )
_ = plt.ylabel('Portfolio Cumulative Return' )

The example above is a perfect example for understanding the behavior

of a portfolio considering four companies. As the plot above demon-
strates, the portfolio went from a cumulative return of approximately 0.4
in 2015 to above 2.0 in 2018. The last part of the graph exemplifies an
interesting fall considering the portfolio (Fig. 1).

Fig. 1 Cumulative returns of the portfolio (Source Elaborated by the author

with information from Yahoo Finance)
186 M. GARITA

As the CAPM model specifies, it is important to calculate the excess

based on the difference of the Risk-Free Rate and the Portfolio Returns.
To choose the risk-free rate concerning a T-Bill should be based on the
duration of the investment. For example, if the portfolio that is being
built is aimed at 10 years then T-Bill that is appropriate is the 10-year
T-Bill.
To calculate the real risk-free rate the process is as follows:

Equation 8: Real Risk

Real Risk − Free Rate = (Risk Free − Rate − Inflation)

At the time consulted for the present chapter, the T-Bill yield for
10 years was 2.53% and the inflation was 1.5%. The result of the real risk-
free rate is 1.03%. If the model is using the real risk-free rate the column
shall be added:

• Add the real risk-free rate column

stocks_return['RF Rate' ] = 0.0103

With the column of the real risk-free rate and the portfolio returns, the
excess can be calculated very easy:

stocks_return['excess' ] = stocks_return['Portfolio' ] - stocks_return['RF Rate' ]

With the information, the only variable missing for the CAPM is the
beta. For this next step, the SPY (S&P 500) will be used:

• Calculating returns for the SPY index

start = datetime.datetime(2014, 1, 2 )
end = datetime.datetime(2019, 1, 1 )

stocks_return['Market' ] = web.DataReader('SPY','yahoo' ,start,end)[ 'Close' ]

stocks_return['Market' ] = (stocks_return['Market' ]/stocks_return['Market' ].

shift(1 ))
stocks_return.head().dropna()
 VALUATION AND RISK MODELS WITH STOCKS 187

The second step is to calculate the excess return of the market compared
with the risk-free rate. This is basic because it standardizes the process
by which the market behaves compared to the risk-free rate and the
portfolio.

• Calculating the excess return of the market

stocks_return['excess market' ] = stocks_return['Market' ] - stocks_return['RF Rate' ]

stocks_return.head().dropna()

To obtain the beta there are different processes, for this example the pro-
cess of the beta equation using covariance and variance will be used. The
equation is as follows:

Equation 9: Beta calculation with covariance

 
Covar Rp , RB
βp =
Var(RB )

βp = Beta of the portfolio

Rp = Return of the portfolio
RB = Return of the benchmark

The next step is to obtain the covariance matrix. From the covariance
matrix the coefficient can be inserted into a variable:

• Covariance Matrix

covariance_matrix = stocks_return[['excess', 'excess market' ]].cov()

covariance_matrix

Covariance Matrix

excess excess market

excess 0.001065 0.000091
excess market 0.000091 0.000070
188 M. GARITA

• Covariance Coefficient

covariance_coefficient = covariance_coefficient = covariance_matrix.

iloc[0, 1]

Once the covariance is obtained the next process is to insert the variance
of the portfolio into the process.

• Calculating the variance

variance_coefficient = stocks_return['excess market' ].var()

variance_coefficient

With the variance and the covariance, the beta can be obtained by using
the formula described before.

• Calculating Beta

beta = covariance_coefficient/variance_coefficient
beta

1.2979841588534764

Considering the process, the beta demonstrates that the portfolio is

more volatile than the market. Another interpretation is that the portfo-
lio is 29% more volatile than the S&P 500 or that for every 1% of move-
ment in the market there will be a 1.29% of rise or fall in the portfolio.

Sharpe Ratio
The Sharpe Ratio was created by William F. Sharpe based on the impor-
tance of understanding the relation between risk and returns. The Sharpe
Ratio is considered one of the rentability and risk ratios.
When the Sharpe Ratio is higher the better it is considered since the
denominator is standard deviation or risk. It is useful when comparing
peers, for example in an exchange-traded fund (ETF) (Hargrave 2019).
 VALUATION AND RISK MODELS WITH STOCKS 189

Equation 10: Sharpe Ratio

Rp − Rf
Sharpe Ratio =
σp
where:
Rp = returns of the portfolio
Rf = risk-free rate
σp = standard deviation of the portfolio excess returns.

For obtaining the Sharpe Ratio the first step is to create the portfolio
using the method that has been used before:

• Installing packages

import numpy as np

import pandas as pd

from pandas_datareader import data as web

import pandas_datareader

import datetime

import matplotlib.pyplot as plt

% matplotlib inline

• Creating the portfolio based on an end and start date

start = datetime.datetime(2018, 1, 2 )
end = datetime.datetime(2019, 4, 1 )

tickers = ['F','FCAU', 'TM' ,]

stocks = pd.DataFrame()
for x in tickers:
stocks[x] = web.DataReader(x, 'yahoo' , start, end)[ 'Close' ]

In the example above the companies that are being used are Ford (F),
Fiat-Chrysler (FCAU) and Toyota Motors (TM). The dates are from
January 1, 2018 to April 1, 2019. For this example, a for loop was used
190 M. GARITA

as an easiest way to understand the process by which the stocks can be

added.

• Analyze correlation between variables

stocks.corr()

Correlation between companies

F FCAU TM
F 1.000000 0.873512 0.837771
FCAU 0.873512 1.000000 0.852925
TM 0.837771 0.852925 1.000000

Concerning the correlation, the portfolio will be strongly correlated

between the securities. This is often a recommendation to diversify the
portfolio and to add companies that are in a different industry.

• Choose weights for the portfolio

portfolio_weights = np.array([0.33, 0.33, 0.34 ])

• Returns of the stocks using percent change

stocks_return = stocks.pct_change(1 ).dropna()

The use of the percent change is an easier approach because it takes the
last return and calculates the change in percentages, which is another
way of calculating the stocks returns without using logarithmic returns.
Mathematically it is not recommended to use logarithmic model to cal-
culate returns (Fig. 2).

• Multiply the returns of the portfolio with each stock’s weight

weighted_returns_portfolio = stocks_return.mul(portfolio_weights, axis = 1 )

• Create a variable for the portfolio by calculating the sum of the returns

stocks_return[ 'Portfolio' ] = weighted_returns_portfolio.sum(axis = 1 ).

dropna()
stocks_return.tail()
 VALUATION AND RISK MODELS WITH STOCKS 191

Fig. 2 Comparing Benchmark and Portfolio (Source Elaborated by the author

with information from Yahoo Finance)

• Add the benchmark by using the S&P 500 and calculating the
returns

start = datetime.datetime(2018, 1, 2 )
end = datetime.datetime(2019, 4, 1 )

stocks_return['Benchmark' ] = web.DataReader('SPY','yahoo' ,start,end)[ 'Close' ]

stocks_return[ 'Benchmark' ] = stocks_return[ 'Benchmark' ].pct_change(1).
dropna()
stocks_return.dropna().tail()

• Obtaining the cumulative returns and plotting the portfolio

CumulativeReturns = ((1 + stocks_return[[ 'Portfolio','Benchmark' ]]).cum-

prod()-1 )
CumulativeReturns.plot(figsize = (16,4 ))
_ = plt.ylabel('Returns' )
_ = plt.title('Comparison - Portfolio vs. Benchmark' )
_ = plt.xlabel('Date' )
plt.show()

• Create a scatterplot to identify the correlation between the portfolio

and the benchmark

plt.scatter(stocks_return['Portfolio' ],stocks_return['Benchmark' ],alpha = 0.80 );

_ = plt.ylabel('Returns' )
_ = plt.title('Correlation - Portfolio vs. Benchmark' )
_ = plt.xlabel('Date' )
192 M. GARITA

• Create a new DataFrame for the portfolio and the benchmark for
calculating the correlation

portfolio_benchmark = pd.concat([stocks_return[ 'Portfolio' ],stocks_

return['Benchmark' ]],axis = 1 ).dropna()
portfolio_benchmark.columns = ['Portfolio',' Benchmark' ]

• Obtain the correlation between the portfolio and the benchmark

correlation = portfolio_benchmark.corr()
correlation

Correlation Between Portfolio and the Benchmark

Portfolio Benchmark
Portfolio 1.000000 0.666916
Benchmark 0.666916 1.000000

• Add risk-free rate based on the benchmark of Treasury Bills chosen

stocks_return['RF Rate' ] = 0.0103.

• Calculate the excess of the portfolio

stocks_return['excess' ] = stocks_return['Portfolio' ] - stocks_return['RF Rate' ]

• Calculate the excess of the benchmark

stocks_return['excess_b' ] = stocks_return['Benchmark' ] - stocks_return['RF Rate' ]

• Calculating the Sharpe Ratio

sharpe_ratio = ((stocks_return['Portfolio' ].mean() - stocks_return['RF Rate' ].

mean()))/stocks_return['Portfolio' ].std()
sharpe_ratio

-0.7134969693218578
 VALUATION AND RISK MODELS WITH STOCKS 193

• Calculating the annual Sharpe Ratio

import math
annual_days = 252

sharpe_ratio_annual = sharpe_ratio * math.sqrt(annual_days).

sharpe_ratio_annual

-11.18589313516733

Considering that the results are negative, the problem lies that the
mean return of the portfolio is smaller than the risk-free rate. As dis-
cussed before, the return of the portfolio is lower than the risk-free rate
exemplifies that the portfolio is not performing above our lowest target
(the RF Rate) and therefore a portfolio is not effective (Fig. 3).
By using highly correlated assets as a portfolio, returns were sacrificed
and the risk was higher because if there is a fall in one security the other
companies will respond in the same way. This is an important lesson
when building a portfolio and not considering the warnings when ana-
lyzing correlation.

Fig. 3 Correlation plot between Portfolio and Benchmark (Source Elaborated

by the author with information from Yahoo Finance)
194 M. GARITA

Traynor Ratio
The ratio was created by John Traynor in 1965 and measure the rentabil-
ity compared to risk. The rule of thumb of the ratio is that an indicator
that is higher is a result of the portfolio management. When analyzing
the Traynor Ratio, if it is negative, the portfolio has underperformed the
risk-free rate.
Since the Traynor Ratio measures returns in excess earned on a riskless
investment compared by using a per market risk, it is useful because it
compares the returns to the risk of the investor (Keaton 2020). As seen
in the following equation, the main difference between a Sharpe Ratio
and a Traynor Ratio is the use of beta.

Equation 11: Traynor Ratio

Rp − Rf
Traynor Ratio =
βp
where:
Rp = returns of the portfolio
Rf = risk-free rate
βp = Beta of the portfolio

For example, the data set that was used for the Sharpe Ratio will be
used. This is important for comparison between the ratios.

• Installing packages

import numpy as np

import pandas as pd

from pandas_datareader import data as web

import pandas_datareader

import datetime

import matplotlib.pyplot as plt

% matplotlib inline
 VALUATION AND RISK MODELS WITH STOCKS 195

• Creating the portfolio based on an end and start date

start = datetime.datetime(2018, 1, 2 )
end = datetime.datetime(2019, 4, 1 )
tickers = ['F','FCAU', 'TM' ,]
stocks = pd.DataFrame()
for x in tickers:
stocks[x] = web.DataReader(x, 'yahoo' , start, end)[ 'Close' ]

• Choose weights for the portfolio

portfolio_weights = np.array([0.33, 0.33, 0.34 ])

• Returns of the stocks using percent change

stocks_return = stocks.pct_change(1 ).dropna()

• Multiply the returns of the portfolio with each stock’s weight

weighted_returns_portfolio = stocks_return.mul(portfolio_weights,
axis = 1 )

• Create a variable for the portfolio by calculating the sum of the

returns

stocks_return[ 'Portfolio' ] = weighted_returns_portfolio.sum(axis = 1 ).

dropna()
stocks_return.tail()

• Add the benchmark by using the S&P 500 and calculating the
returns

start = datetime.datetime(2018, 1, 2 )
end = datetime.datetime(2019, 4, 1 )

stocks_return['Benchmark' ] = web.DataReader('SPY','yahoo' ,start,end)[ 'Close' ]

stocks_return[ 'Benchmark' ] = stocks_return[ 'Benchmark' ].pct_change(1).
dropna()
stocks_return.dropna().tail()
196 M. GARITA

• Obtaining the cumulative returns and plotting the portfolio

CumulativeReturns = ((1 + stocks_return[[ 'Portfolio','Benchmark' ]]).cum-

prod()-1 )
CumulativeReturns.plot(figsize = (16,4 ))
_ = plt.ylabel('Returns' )
_ = plt.title('Comparison - Portfolio vs. Benchmark' )
_ = plt.xlabel('Date' )
plt.show()

• Add risk-free rate based on the benchmark of Treasury Bills chosen

stocks_return['RF Rate' ] = 0.0103

• Calculate the covariance of the stocks

covariance = stocks_return.cov() * 252

covariance

F FCAU TM Portfolio Benchmark

F 0.081996 0.056534 0.024013 0.053879 0.023830
FCAU 0.056534 0.167189 0.036997 0.086408 0.035823
TM 0.024013 0.036997 0.035576 0.032229 0.019736
Portfolio 0.053879 0.086408 0.032229 0.057070 0.026396
Benchmark 0.023830 0.035823 0.019736 0.026396 0.027260

• Calculate the covariance between the market and the portfolio

covariance_market = covariance.iloc[3,4 ]
covariance_market

0.026395748767295644

• Calculate the variance of the benchmark

market_variance = stocks_return['Benchmark' ].var() * 252

market_variance
 VALUATION AND RISK MODELS WITH STOCKS 197

Fig. 4 Comparsion between portfolio and benchmark (Source Elaborated by

the author with information from Yahoo Finance)

• Calculate the beta of the portfolio

portfolio_beta = covariance_market / market_variance

portfolio_beta

0.96828357625053052

traynor_ratio = ((stocks_return[‘Portfolio’].mean() - stocks_return['RF Rate' ].

mean()))/portfolio_beta
traynor_ratio

-0.011269427871314531

As a result, the Traynor Ratio is negative, which means that the port-
folio is not performing better than the risk-free rate. The result of the
Sharpe Ratio was negative also, so it complements the information given
by Traynor. The main difference between the Sharpe and the Traynor
ratio is that it compares with the beta and not the volatility (Fig. 4).

Jensen’s Measure
The Jensen’s Measure also known as alpha was created in 1968 with the
purpose of measuring the relationship between the return of the portfo-
lio in comparison with another portfolio return with the same risk, same
reference market and under the same parameters (Chen 2019).
198 M. GARITA

Equation 12: Jensen’s measure

 
Alpha (α) = Rp − (Rf + βp Rm − Rp

Rp = returns of the portfolio

Rf = risk-free rate
βp = Beta of the portfolio
Rm = return of the market.

• Installing packages

import numpy as np

import pandas as pd

from pandas_datareader import data as web

import pandas_datareader

import datetime

import matplotlib.pyplot as plt

% matplotlib inline

• Creating the portfolio based on an end and start date

start = datetime.datetime(2018, 1, 2 )
end = datetime.datetime(2019, 4, 1 )

tickers = ['F','FCAU', 'TM',]

stocks = pd.DataFrame()
for x in tickers:
stocks[x] = web.DataReader(x, 'yahoo' , start, end)[ 'Close' ]

• Choose weights for the portfolio

portfolio_weights = np.array([0.33, 0.33, 0.34 ])

• Returns of the stocks using percent change

stocks_return = stocks.pct_change(1 ).dropna()

 VALUATION AND RISK MODELS WITH STOCKS 199

• Multiply the returns of the portfolio with each stock’s weight

weighted_returns_portfolio = stocks_return.mul(portfolio_weights,
axis = 1 )

• Create a variable for the portfolio by calculating the sum of the

returns

stocks_return[ 'Portfolio' ] = weighted_returns_portfolio.sum(axis = 1 ).

dropna()
stocks_return.tail()

• Add the benchmark by using the S&P 500 and calculating the
returns

start = datetime.datetime(2018, 1, 2 )
end = datetime.datetime(2019, 4, 1 )

stocks_return['Benchmark' ] = web.DataReader('SPY','yahoo' ,start,end)[ 'Close' ]

stocks_return[ 'Benchmark' ] = stocks_return[ 'Benchmark' ].pct_change(1).
dropna()
stocks_return.dropna().tail()

• Add risk-free rate based on the benchmark of Treasury Bills chosen

stocks_return['RF Rate' ] = 0.0103

• Calculate the covariance of the stocks

covariance = stocks_return.cov() * 252

covariance

F FCAU TM Portfolio Benchmark

F 0.081996 0.056534 0.024013 0.053879 0.023830
FCAU 0.056534 0.167189 0.036997 0.086408 0.035823
TM 0.024013 0.036997 0.035576 0.032229 0.019736
Portfolio 0.053879 0.086408 0.032229 0.057070 0.026396
Benchmark 0.023830 0.035823 0.019736 0.026396 0.027260
200 M. GARITA

• Choose the covariance between the market and the portfolio

covariance_market = covariance.iloc[3,4 ]
covariance_market

0.026395748767295644

• Calculate the variance of the benchmark

market_variance = stocks_return['Benchmark' ].var() * 252

market_variance

• Calculate the beta of the portfolio

portfolio_beta = covariance_market / market_variance

portfolio_beta

0.96828357625053052

• Calculate the return of the portfolio

portfolio_return = stocks_return['Portfolio' ].mean()

• Calculate the risk-free rate

risk_free_rate = stocks_return['RF Rate' ].mean()

• Calculate the Alpha

alpha = portfolio_return - (risk_free_rate + portfolio_beta *(portfolio_

return - risk_free_rate)).
alpha

-0.00034608967689839253

The alpha is negative and so is the Traynor and the Sharpe Ratio. This
is a perfect example of a portfolio that is performing negatively when
compared to the risk-free rate. For decision making with alpha the fol-
lowing table is important (Table 6):
 VALUATION AND RISK MODELS WITH STOCKS 201

Table 6 Alpha decision making

(α) > 0 the portfolio has gained value

(α) < 0 the portfolio has lost value

Information Ratio
The information ratio analyzes the excess of the return when compar-
ing the portfolio without risk with the one supported by the investment.
The effect that the ratio measures is how the portfolio deviates from the
benchmark (Murphy 2019) The name is based on the consideration that
the manager of the portfolio has special information and therefore, he
will out beat the benchmark. The formula is as follows:
(Rp − Rm )
Information Ratio =
TE
TE = standard deviation of the difference between the portfolio and the
benchmark
Rp = Porfolio Return
βp = Beta of the portfolio
Rm = return of the market

• Installing packages

import numpy as np

import pandas as pd

from pandas_datareader import data as web

import pandas_datareader

import datetime

import matplotlib.pyplot as plt

% matplotlib inline
202 M. GARITA

• Creating the portfolio based on an end and start date

start = datetime.datetime(2018, 1, 2 )
end = datetime.datetime(2019, 4, 1 )

tickers = ['F','FCAU', 'TM',]

stocks = pd.DataFrame()
for x in tickers:
stocks[x] = web.DataReader(x, 'yahoo', start, end)[ 'Close' ]

• Choose weights for the portfolio

portfolio_weights = np.array([0.33, 0.33, 0.34 ])

Returns of the stocks using percent change

stocks_return = stocks.pct_change(1 ).dropna()

• Multiply the returns of the portfolio with each stock’s weight

weighted_returns_portfolio = stocks_return.mul(portfolio_weights,
axis = 1 )

• Create a variable for the portfolio by calculating the sum of the

returns.

stocks_return[ 'Portfolio' ] = weighted_returns_portfolio.sum(axis = 1 ).

dropna()
stocks_return.tail()

• Add the benchmark by using the S&P 500 and calculating the
returns

start = datetime.datetime(2018, 1, 2 )
end = datetime.datetime(2019, 4, 1 )

stocks_return['Benchmark' ] = web.DataReader('SPY','yahoo' ,start,end)['Close' ]

stocks_return[ 'Benchmark' ] = stocks_return[ 'Benchmark' ].pct_change(1).
dropna()
stocks_return.dropna().tail()
 VALUATION AND RISK MODELS WITH STOCKS 203

• Add risk-free rate based on the benchmark of Treasury Bills chosen

stocks_return['RF Rate' ] = 0.0103

• Calculate the covariance of the stocks

covariance = stocks_return.cov() * 252

covariance

F FCAU TM Portfolio Benchmark

F 0.081996 0.056534 0.024013 0.053879 0.023830
FCAU 0.056534 0.167189 0.036997 0.086408 0.035823
TM 0.024013 0.036997 0.035576 0.032229 0.019736
Portfolio 0.053879 0.086408 0.032229 0.057070 0.026396
Benchmark 0.023830 0.035823 0.019736 0.026396 0.027260

• Choose the covariance between the market and the portfolio

covariance_market = covariance.iloc[3,4 ]
covariance_market

0.026395748767295644

• Calculate the variance of the benchmark

market_variance = stocks_return['Benchmark' ].var() * 252

market_variance

• Calculate the beta of the portfolio

portfolio_beta = covariance_market / market_variance

portfolio_beta

0.96828357625053052

• Calculate the return of the portfolio

portfolio_return = stocks_return['Portfolio' ].mean()

204 M. GARITA

• Calculate the return of the benchmark

portfolio_return = stocks_return['Benchmark' ].mean()

• Difference between the return of the portfolio and the benchmark

difference_benchmark_portfolio = stocks_return[‘Portfolio’] - stocks_

return['Benchmark' ]

• Calculate the tracking error

tracking_error = difference_benchmark_portfolio.std()

Calculate the information

information_ratio = ((portfolio_return - benchmark_return))/ tracking_error

information_ratio

-0.07703212677308867

The result demonstrates that there is no excess between the returns

of the portfolio and the benchmark leading to infer that the bench-
mark has outperformed the portfolio. As a conclusion concerning the
ratios, the portfolio has performed worse than the risk-free rate and the
benchmark.

• Applying the knowledge to an investment

The following exercise is based on an investment of USD 300,000 in the

following securities with a specific allocation:
– Amazon 30% allocation—USD 90,0000
– Ford 20% allocation—USD 60,000
– Citi 30% allocation—USD 90,000
– McDonalds 20%—USD 60,000
 VALUATION AND RISK MODELS WITH STOCKS 205

• Installing packages

import numpy as np

import pandas as pd

from pandas_datareader import data as web

import pandas_datareader

import datetime

import matplotlib.pyplot as plt

% matplotlib inline

• Retrieving the information concerning the securities

start = datetime.datetime(2019, 1, 2 )
end = datetime.datetime(2020, 4, 1 )

Amazon = web.DataReader('AMZN', 'yahoo' , start, end)

Ford = web.DataReader('F', 'yahoo' , start, end)
Citi = web.DataReader('C', 'yahoo' , start, end)
McDonalds = web.DataReader('MCD', 'yahoo' , start, end)

• Calculating the returns for the portfolio using closing price

for securities in (Amazon,Ford,Citi,McDonalds):

securities['Return' ] = securities['Close' ] /securities.iloc[0 ][ 'Close' ]

• Calculating the returns for the portfolio using closing price

for securities , allocation in zip((Amazon,Ford,Citi,MacDonalds),[.3,.2,.3,.2]):

securities['Allocation' ] = securities['Return' ]*allocation.

For the first time the function zip()5 is used in the present book. The
reason for this is that it allows an iteration between the variables. In this
case the allocation and the return.

5 For more information visit: https://realpython.com/python-zip-function/

206 M. GARITA

• Establishing the investment based on the investment of USD

300,000

(Amazon,Ford,Citi,MacDonalds):
for securities in
securities['Investment' ] = securities['Allocation' ]* 300000

securities.tail()

High Low Open Close Volume Adj Close Return Allocation Investment

Date

2020–03-26 170.929993 161.000000 163.990005 167.350006 8,259,900.0 167.350006 0.950528 0.190106 57,031.696612

2020–03-27 169.740005 159.220001 162.779999 164.009995 6,441,400.0 164.009995 0.931557 0.186311 55,893.444319

2020–03-30 170.309998 163.570007 164.919998 168.130005 5,621,700.0 168.130005 0.954959 0.190992 57,297.514670

2020–03-31 169.509995 165.000000 166.839996 165.350006 4,519,900.0 165.350006 0.939169 0.187834 56,350.110779

2020–04-01 161.440002 156.350006 160.220001 158.169998 4,668,900.0 158.169998 0.898387 0.179677 53,903.214937

• Investment per security chosen

all_investments =
[Amazon['Investment'],Ford['Investment'],Citi['Investment'],MacDonalds['Investment']]
value_of_portfolio = pd.concat(all_investments,axis = 1)
value_of_portfolio.columns = [
'Amazon Investment','Ford Investment','Citi Investment','McDonalds Investment' ]
value_of_portfolio.head()

• Analyzing the positions

value_of_portfolio[‘Total Investment’] = value_of_portfolio.sum(axis = 1)

value_of_portfolio.tail()

Amazon Ford Investment Citi Investment McDonalds Total Investment

Investment Investment
Date
2020–03-26 76,230.987013 39,873.417240 103,164.583988 57,031.696612 276,300.684853
2020–03-27 74,071.714653 39,417.721478 98,187.932530 55,893.444319 267,570.812979
2020–03-30 76,560.782193 38,202.532778 98,815.623769 57,297.514670 270,876.453410
2020–03-31 76,006.053986 36,683.543282 94,421.819299 56,350.110779 263,461.527346
2020–04-01 74,367.984970 33,417.721840 86,329.159424 53,903.214937 248,018.081171
 VALUATION AND RISK MODELS WITH STOCKS 207

• Plotting the value of the portfolio

value_of_portfolio['Total Investment' ].plot(figsize = (10,8 )).

_ = plt.xlabel('Investment Performance' )
_ = plt.ylabel('Date' )
_ = plt.title('Portfolio investment' )
plt.legend();

As it can observed in Fig. 23 in Chapter 6, the portfolio began with

the investment of USD 300,000 and reached its highest value at USD
369,373 and its lowest value at USD 247,946.82. To calculate this val-
ues, it can be done with the min() and max() values (Fig. 5).

value_of_portfolio[‘Total Investment’].max()
396373.7772855786

value_of_portfolio[‘Total Investment’].min()
247946.82954672351
In [118]:

Fig. 5 Total position of the portfolio (Source Elaborated by the author with
information from Yahoo Finance)
208 M. GARITA

• Plotting the behavior of each security

value_of_portfolio.drop('Total Investment' ,axis = 1).plot(figsize = (10,8 ));

In Fig. 24 in Chapter 6, it can be observed that the only security that

is gaining is Amazon, McDonald’s, Citi and Ford have had a fall losing
value in the portfolio (Fig. 6).

• Calculating the Daily Returns

value_of_portfolio[ 'Daily Returns' ] = value_of_portfolio[ 'Total Investment' ].

pct_change(1)

• Calculating the mean of the Daily Returns

value_of_portfolio['Daily Returns' ].mean()

0.0003811570091341308

• Calculating the standard deviation of the Daily Returns

value_of_portfolio['Daily Returns' ].std()

0.02105377310080987

Fig. 6 Position behavior on each security (Source Elaborated by the author with
information from Yahoo Finance)
 VALUATION AND RISK MODELS WITH STOCKS 209

• Calculating the cumulative returns of the portfolio

cumulative_return = 100 * (value_of_portfolio['Total Investment' ][-1]/value_

of_portfolio['Total Investment' ][0]-1)
12.126738733409015

The loss of the portfolio is of negative 12%. This can be analyzed

based on the cumulative returns since the beginning of the portfolio.

• Value of the portfolio to date

value_of_portfolio['Total Investment' ][-1]

263619.78379977297

The portfolio of USD 300,000 has fallen to USD 263,619.78 which

is equivalent to a 12.12% fall based on the Covid-19 crisis.

References
Bryant, Bradley James. 2020. How to Calculate Portfolio Value. n.d.
Accessed January 3, 2020. https://www.sapling.com/5872650/
calculate-portfolio-value.
Chen, James. 2019. Jensen's Measure. 21 November. Accessed December 20,
2019. https://www.investopedia.com/terms/j/jensensmeasure.asp.
Chen, James. 2019. Financial Risk. 15 June. Accessed August 20, 2019.
https://www.investopedia.com/terms/f/financialrisk.asp.
David W. Mullins, Jr. 1982. Does the Capital Asset Pricing Model Work?
n.d. January. Accessed August 13, 2019. https://hbr.org/1982/01/
does-the-capital-asset-pricing-model-work.
Fontinelle, Amy. 2019. Systematic Risk. 30 September. Accessed October 1,
2019. https://www.investopedia.com/terms/s/systematicrisk.asp.
Hargrave, Marshall. 2019. Sharpe Ratio. 17 March. Accessed April 1, 2019.
https://www.investopedia.com/terms/s/sharperatio.asp.
Keaton, Will. 2020. Treynor Ratio. 22 March. Accessed Apri 1, 2020. https://
www.investopedia.com/terms/t/treynorratio.asp.
Mitra, Gautam, and Leela Mitra. 2011. The Handbook of News Analytics in
Finance. London: Wiley.
Murphy, Chris. 2019. Information Ratio – IR. 10 January. Accessed February
20, 2019. https://www.investopedia.com/terms/i/informationratio.asp.
 Value at Risk

Abstract Focusing on the creation of portfolios for investment, this chap-

ter aims to understand the risks of the portfolio through methods such as
the Value at Risk (VaR) to determine the possible loss or gain of a portfo-
lio. This chapter is based on an investor view and the process for executing
decisions that create profitable portfolios in the short and long run.

Keywords Risk · Portfolios · VaR · Backtesting

The concept of Value at Risk (VaR) is one of the most interesting in

finance because it analyzes the maximum loss that a portfolio may have
(Damodaran 2018). This is another measure of risk that deserves to be
separated from portfolio and risk because of the difference that it has
with the ratios (Sharpe, Traynor, Information and Jensen) in the previ-
ous chapter. To summarize the VaR, it gives the worst loss on a certain
time horizon based on the confidence level assigned to the model.

Historical VaR(95)
Since the VaR is based on the confidence level, it may have different
results based on a 65%, 90%, 95% or any other confidence interval. The
following example is Historical VaR(95), meaning that the confidence
interval will be at a 95%.

© The Author(s), under exclusive license to Springer 211

Nature Switzerland AG 2021
M. Garita, Applied Quantitative Finance,
https://doi.org/10.1007/978-3-030-29141-9_8
212 M. GARITA

• Install packages

LPSRUWQXPS\DVQS
LPSRUWSDQGDVDVSG
IURPSDQGDVBGDWDUHDGHULPSRUWGDWDDVZHE
LPSRUWSDQGDVBGDWDUHDGHU
LPSRUWGDWHWLPH
LPSRUWPDWSORWOLES\SORWDVSOW
PDWSORWOLELQOLQH

• Choose the portfolio

VWDUW GDWHWLPHGDWHWLPH 

HQG GDWHWLPHGDWHWLPH 

WLFNHUV > $$3/  :07  70  .2  %$ @

VWRFNV SG'DWD)UDPH
IRU[LQWLFNHUV
VWRFNV>[@ ZHE'DWD5HDGHU [ \DKRR VWDUWHQG > &ORVH @

VWRFNVWDLO

• Calculate the returns

stocks_return = (stocks/stocks.shift(1))−1
stocks_return.tail()

• Assign random portfolio weights that sum to one (1)

• portfolio_weights = np.array(np.random.random(5))

portfolio_weights
portfolio_weights = portfolio_weights/np.sum(portfolio_weights)
portfolio_weights
 VALUE AT RISK 213

This step is interesting because, in the Portfolio and Risk chapter,

the purpose was to assign the same return to each of the stocks. In this
case the np.random.random creates weights for the five (5) stocks but
it often gives a number less or higher than 100%. Therefore it has to be
balanced by dividing the weights in the sum to obtain a portfolio that
sums 100%.

• Multiply the portfolio with the stocks

weighted_returns_portfolio = stocks_return.mul(portfolio_weights,
axis = 1)

• Convert returns to percentages and drop the missing values

stocks_return['Portfolio'] = weighted_returns_portfolio.sum(axis=1).dropna()
stocks_return['Portfolio'] = stocks_return['Portfolio'] * 100

• Calculate the VaR95

var95 = np.percentile(stocks_return['Portfolio'], 5) * 100

var95
−1.6739577270187669

Based on the historical returns of the portfolio at a 95% confidence inter-

val, the worst loss is a 1.67% loss, therefore the result is negative.

Historical VaR(99)
For computing the Historical VaR at a 99% confidence level the only
change that has to be done is in the last part of the script, changing the
np.percentile to 1, which means the 1%.

var99 = np.percentile(stocks_return['Portfolio'], 5) * 100

var99

-2.5793928700853099
214 M. GARITA

At a 99% confidence level the worst loss is 2.58% with the portfolio.
Clearly the VaR is higher given that the confidence level is lower. This
is rational and therefore it helps understand the process by which the
VaR works, given that a higher confidence level will give a higher per-
centage of loss and a lower confidence level will give a lower percentage
of loss.

VaR for the Next 10 Days

One of the most important aspects of calculating a VaR is to calculate the
effect on the investment in terms of money. As far, the VaR model has
centered on the percentage loss, but for the next example the process is
to analyze the VaR if USD 1 million is invested. For example, the same
data set will be used.

• Install packages

#installing packages

• Installing packages

import numpy as np
import pandas as pd
from pandas_datareader import data as web
import pandas_datareader
import datetime
import matplotlib.pyplot as plt
%matplotlib inline

• Choose the portfolio

 VALUE AT RISK 215

VWDUW GDWHWLPHGDWHWLPH 

HQG GDWHWLPHGDWHWLPH 

WLFNHUV > $$3/  :07  70  .2  %$ @

VWRFNV SG'DWD)UDPH
IRU[LQWLFNHUV
VWRFNV>[@ ZHE'DWD5HDGHU [ \DKRR VWDUWHQG > &ORVH @

VWRFNVWDLO

• Calculate the returns

stocks_return = (stocks/stocks.shift(1))−1
stocks_return.tail()

• Assign random portfolio weights that sum to one (1)

portfolio_weights = np.array(np.random.random(5))
portfolio_weights
portfolio_weights = portfolio_weights/np.sum(portfolio_weights)
portfolio_weights

• Multiply the portfolio with the stocks

weighted_returns_portfolio = stocks_return.mul(portfolio_weights,
axis = 1)

• Calculate the returns based on the weights

VWRFNVBUHWXUQ> 3RUWIROLR @ ZHLJKWHGBUHWXUQVBSRUWIROLRVXP D[LV  GURSQD

• Determine the average (mu) of the returns

PX VWRFNVBUHWXUQ> 3RUWIROLR @PHDQ

216 M. GARITA

• Determine the standard deviation (sigma) of the returns

VLJPD VWRFNVBUHWXUQ> 3RUWIROLR @VWG

• Assign a confidence level to the VaR (99% for this example)

confidence = 0.99

• Calculate the alpha

alpha = norm.ppf(1−confidence)

For this example, the norm.ppf is being used, the reason for this is that
it determines the probability density function of one (1) minus the confi-
dence interval. This is useful because it determines the probability of the
VaR. It is a similar process to the np.percentile.

• Create a position

position = 1e6

The position is the investment on the portfolio. Since the portfolio

was created, in this case the investment is USD 1 million. The interesting
aspect of using 1e6 for a million is to include a complex number struc-
ture that is easier to write. The other choice would have been to write
the 1,000,000.

• Calculate the VaR

Equation 1: Value at Risk - position

VaR = position * (µ − σ * α)
µ = mean of the returns of the portfolio
σ = standard deviation of the returns of the portfolio
α = Probability density function of the 1%

var = position*(mu−sigma*alpha)
var
27088.745452792264
 VALUE AT RISK 217

If the investment in the portfolio was of USD 1,000,000, the worst

loss at a 99% confidence interval can be of USD 27,088.75. The next
step is to obtain the VaR for the next 10 days, trying to identify what will
be the loss of the portfolio.

• Create a variable for 10 days

days = 10

• Determine the worst loss for the next 10 days


VaR for 10 days = position ∗ (µ ∗ days − σ ∗ α ∗ days)

var_10_days = position *(mu*days−sigma*alpha*np.sqrt(days))

var_10_days

88644.949585607217

The worst loss for the next 10 days based on the portfolio that
depends on the stocks that have been chosen and the weights of the
stocks, could be of USD 88,644.95 or approximately 88.64% of the total
investment. Consider that this effect is at a 99% confidence interval. If
the example had been done with a 95% of confidence interval, the result
would have been as follows:

• Assign a confidence interval of 95%

confidence = 0.95

• Obtain the alpha

alpha = norm.ppf(1−confidence)

• Determine the worst loss for the next 10 days

var_10_days = position *(mu*days-sigma*alpha*np.sqrt(days))

var_10_days

63954.684614643818
218 M. GARITA

The result is a worst loss much smaller than the one determined at a
99% confidence level. In this example the loss is approximately 63.95%
of the total investment. The result also varies if the days are reduced, for
example to 5 days.

• Worst loss for the next 5 days at a 95% confidence level

days_2 = 5

var_5_days = position *(mu*days_2−sigma*alpha*np.sqrt(days))

var_5_days

61773.537991536054

Historical Drawdown
A historical drawdown is often included with the VaR because it analyzes
the decline in the specific period that the portfolio is being analyzed and
based on the cumulative growth analyzes the peak and therefore the fall
or drawdown of the portfolio (Mitchell 2019). The process is similar to
the VaR but it uses the negative returns of the portfolio.

• Installing packages

LPSRUWQXPS\DVQS
LPSRUWSDQGDVDVSG
IURPSDQGDVBGDWDUHDGHULPSRUWGDWDDVZHE
LPSRUWSDQGDVBGDWDUHDGHU
LPSRUWGDWHWLPH
LPSRUWPDWSORWOLES\SORWDVSOW
PDWSORWOLELQOLQH
 VALUE AT RISK 219

• Choose the portfolio

VWDUW GDWHWLPHGDWHWLPH 

HQG GDWHWLPHGDWHWLPH 

WLFNHUV > $$3/  :07  70  .2  %$ @

VWRFNV SG'DWD)UDPH
IRU[LQWLFNHUV
VWRFNV>[@ ZHE'DWD5HDGHU [ \DKRR VWDUWHQG > &ORVH @

VWRFNVWDLO

• Calculate the returns

stocks_return = (stocks/stocks.shift(1))−1
stocks_return.tail()

• Assign random portfolio weights that sum to one (1)

portfolio_weights= np.array(np.random.random(5))
portfolio_weights
portfolio_weights = portfolio_weights/np.sum(portfolio_weights)
portfolio_weights

• Multiply the portfolio with the stocks

weighted_returns_portfolio = stocks_return.mul(portfolio_weights,
axis = 1)

• Calculate the returns based on the weights

VWRFNVBUHWXUQ> 3RUWIROLR @ ZHLJKWHGBUHWXUQVBSRUWIROLRVXP D[LV  GURSQD

220 M. GARITA

• Calculate the cumulative returns

&XPXODWLYH5HWXUQV  VWRFNVBUHWXUQ>3RUWIROLR@ FXPSURG 

• Plot the cumulative returns (Fig. 1)

&XPXODWLYH5HWXUQVSORW
B SOW[ODEHO 'DWHV V

B SOW\ODEHO 5HWXUQV

B SOWWLWOH &XPXODWLYH5HWXUQV3RUWIROLR

SOWVKRZ

Fig. 1 Cumulative Return of the portfolio (Source Elaborated by the author

with information from Yahoo Finance)
 VALUE AT RISK 221

• Determine the running maximum

running_maximum = np.maximum.accumulate(CumulativeReturns)
running_maximum.tail()

The running maximum is taking the maximum of the cumula-

tive returns and accumulates it so that there is a constant return. This
is because the drawdown formula divides the cumulative return of the
portfolio into the running maximum. The np.maximum.accumulate
accumulates all the elements.

• Establish that the running maximum should not go below zero

running_maximum = running_maximum < 1

• Calculate the drawdown

portfolio_drawdown = (CumulativeReturns)/running_max − 1

• Plot the drawdown (Fig. 2)

GUDZGRZQSORW

B SOW[ODEHO 'DWHV

B SOW\ODEHO 5HWXUQV

B SOWWLWOH 'URZGRZQ 3RUWIROLR

SOWVKRZ

As can be seen in the drawdown the returns are negative and it

demonstrates that the worst loss is between the end of 2018 and
February 2019. It is an interesting approach based on the drawdown to
observe how the portfolio could behave to a VaR. In this case the worst
loss is 1.4% being a very stable portfolio.
222 M. GARITA

Fig. 2 Drawdown of the portfolio (Source Elaborated by the author with infor-
mation from Yahoo Finance)

Wrapping Up the Book—Understanding Performance

The book has centered itself on stock trading, and the purpose of this
chapter is to understand performance by using the f.fn() package for cre-
ating a report. I believe this is one of the most interesting tools that can
be used for a quick decision making on investment that does not involve
a graphical decision. This can be applied to a portfolio and individual
stocks. The application of the portfolio is as follows:

Portfolio Performance using f.fn()

– Import libraries
LPSRUWIIQ
LPSRUWQXPS\DVQS
LPSRUWSDQGDVDVSG
IURPSDQGDVBGDWDUHDGHULPSRUWGDWDDVZHE
LPSRUWSDQGDVBGDWDUHDGHU
LPSRUWGDWHWLPH
LPSRUWPDWSORWOLES\SORWDVSOW
PDWSORWOLELQOLQH
 VALUE AT RISK 223

– Select stocks
VWDUW GDWHWLPHGDWHWLPH 
HQG GDWHWLPHGDWHWLPH 

WLFNHUV > =0  $0=1  '2&8  3721 @

VWRFNV SG'DWD)UDPH
IRU[LQWLFNHUV
VWRFNV>[@ ZHE'DWD5HDGHU [ \DKRR VWDUWHQG > &ORVH @

– Calculate returns using f.fn ()

stocks_return = stocks.to_returns().dropna()
stocks_return.tail()

– Calculating mean variance returns with f.fn ( )

PHDQBYDULDQFHBZHLJKWV 
VWRFNVBUHWXUQFDOFBPHDQBYDUBZHLJKWV DVBIRUPDW 
PHDQBYDULDQFHBZHLJKWV

=0
$0=1
'2&8
3721
GW\SHREMHFW

– Applying the mean variance returns to portfolio weights,

SRUWIROLRBZHLJKWV  
SRUWIROLRBZHLJKWV

– Creating the portfolio based on the returns

SRUWIROLRBZHLJKWV SRUWIROLRBZHLJKWVQSVXP SRUWIROLRBZHLJKWV
ZHLJKWHGBUHWXUQVBSRUWIROLR VWRFNVBUHWXUQPXO SRUWIROLRBZHLJKWVD[LV

VWRFNVBUHWXUQ> 3RUWIROLR @ 
ZHLJKWHGBUHWXUQVBSRUWIROLRVXP D[LV  GURSQD
VWRFNVBUHWXUQWDLO
224 M. GARITA

ZM AMZN DOCU PTON BA Portfolio

Date
2020-12-23 −0.061418 −0.006627 −0.030008 0.009615 0.004159 −0.016579
2020-12-24 −0.022689 −0.003949 0.003606 −0.000246 −0.011562 −0.003977
2020-12-28 −0.063385 0.035071 −0.064222 −0.064774 −0.004881 −0.043309
2020-12-29 0.006716 0.011584 −0.004537 −0.013668 0.000740 −0.002385
2020-12-30 −0.000989 −0.010882 −0.009905 0.032378 0.001942 0.006417

– Calculate performance

performance = stocks_return.calc_stats()

– Display performance

performance.display()

Stat ZM AMZN DOCU PTON Portfol

Start 2019-09-27 2019-09-27 2019-09-27 2019-09-27 2019-09-27
End 2020-12-30 2020-12-30 2020-12-30 2020-12-30 2020-12-30
Risk-free rate 0.00% 0.00% 0.00% 0.00% 0.00%
Total Return −97.87% 31.57% −64.69% −260.40% −126.09%
Daily Sharpe −1.50 −1.50 0.07 - −0.45
Daily Sortino −1.51 −1.55 0.28 inf −1.07
CAGR −95.30% 24.34% −56.24% - -
Max Drawdown −1328.75% −313.08% −200.92% −398.81% −899.28
Calmar Ratio −0.07 0.08 −0.28 - -
MTD −106.91% 27.37% −322.50% −52.12% −75.40%
3m −109.99% −988.87% −191.30% −364.48% 1356.52%
6m −104.94% −137.18% −414.89% 230.78% −55.58%
YTD −105.29% −2215.67% −302.92% 26.28% −53.76%
1Y −143.96% −11.19% 15.13% 26.70% 41.04%
3Y (ann.) −95.30% 24.34% −56.24% - -
5Y (ann.) - - - - -
10Y (ann.) - - - - -
Since Incep. (ann.) −95.30% 24.34% −56.24% - -
Daily Sharpe −1.50 −1.50 0.07 - −0.45
Daily Sortino −1.51 −1.55 0.28 inf −1.07
Daily Mean (ann.) −114,734.65% −76,805.36% 1704.15% inf% −11,126.29%
Daily Vol (ann.) 76,370.55% 51,048.23% 22,749.18% - 24,907.54%
Daily Skew −16.67 −13.99 13.26 - 8.40
Daily Kurt 288.21 224.97 213.60 - 108.93
Best Day 3995.93% 8524.06% 23,064.40% inf% 20,939.52%
Worst Day −83,830.58% −52,654.19% −3831.82% −5624.22% −5896.52%
Monthly Sharpe −1.32 0.66 −0.98 −0.50 −1.31
Monthly Sortino −1.79 2.25 −1.53 −0.62 −1.31
Monthly Mean (ann.) −1280.15% 9042.18% −1050.99% −1291.43% −14,979.80%
Monthly Vol (ann.) 972.34% 13,779.63% 1077.20% 2606.03% 11,440.89%
Monthly Skew −0.18 3.02 0.45 −1.56 −3.23
Monthly Kurt 0.81 11.13 1.07 7.23 10.79
 VALUE AT RISK 225

Stat ZM AMZN DOCU PTON Portfol

Best Month 451.85% 14,246.47% 615.58% 1409.53% 113.76%
Worst Month −693.26% −4548.17% −591.84% −2380.16% −12,427.53%
Yearly Sharpe - - - - -
Yearly Sortino - - - - -
Yearly Mean −105.29% −2215.67% −302.92% 26.28% −53.76%
Yearly Vol - - - - -
Yearly Skew - - - - -
Yearly Kurt - - - - -
Best Year −105.29% −2215.67% −302.92% 26.28% −53.76%
Worst Year −105.29% −2215.67% −302.92% 26.28% −53.76%
Avg. Drawdown −345.25% −201.60% −167.47% −190.98% −267.83%
Avg. Drawdown Days 64.29 63.57 63.71 55.75 49.11
Avg. Up Month 213.22% 2862.18% 285.41% 690.44% 52.58%
Avg. Down Month −223.00% −652.26% −223.22% −307.13% −1573.54%
Win Year % 0.00% 0.00% 0.00% 1000.00% 0.00%
Win 12 m % 20.00% 40.00% 60.00% 40.00% 40.00%

The result of the portfolio, during the uncertainty of 2020 with a

deplorable negative 126.09% of loss, a drawdown of negative 267.83%
with an average of 49.11 days. The best month of the portfolio created a
113.76% but the implication of a worse month surpasses that. With this
information corrections and backtesting can be created for better perfor-
mance when analyzing the data.
If the data wants to be seen with only one stock, the process is similar
as when handling the DataFrame.

performance = stocks_return.calc_stats()

start 2019-09-27 00:00:00

end 2020-12-30 00:00:00
rf 0
total_return −1.26094
cagr NaN
max_drawdown −8.99284
calmar NaN
mtd −0.754048
three_month 13.5652
six_month −0.555818
ytd −0.53764
one_year 0.410417
three_year NaN
five_year NaN
ten_year NaN
incep NaN
daily_sharpe −0.446704
daily_sortino −1.06647
226 M. GARITA

daily_mean −111.263
daily_vol 249.075
daily_skew 8.39942
daily_kurt 108.93
best_day 209.395
worst_day −58.9652
monthly_sharpe −1.30932
monthly_sortino −1.31108
monthly_mean −149.798
monthly_vol 114.409
monthly_skew −3.23038
monthly_kurt 10.7853
best_month 1.1376
worst_month −124.275
yearly_sharpe NaN
yearly_sortino NaN
yearly_mean −0.53764
yearly_vol NaN
yearly_skew NaN
yearly_kurt NaN
best_year −0.53764
worst_year −0.53764
avg_drawdown −2.67834
avg_drawdown_days 49.1111
avg_up_month 0.525848
avg_down_month −15.7354
win_year_perc 0
twelve_month_win_perc 0.4

Fund Performance using f.fn()

As seen before, there are other analyses using performance that can
be done, when analyzing the prices of a stock or in this case, a fund.
Although funds have not been discussed in the book, it is important to
understand that the statistical methods are similar, the interpretation
changes regarding the instrument. The process is as follow:

– Importing libraries
LPSRUWIIQ
LPSRUWQXPS\DVQS
LPSRUWSDQGDVDVSG
IURPSDQGDVBGDWDUHDGHULPSRUWGDWDDVZHE
LPSRUWSDQGDVBGDWDUHDGHU
LPSRUWGDWHWLPH
LPSRUWPDWSORWOLES\SORWDVSOW
PDWSORWOLELQOLQH
 VALUE AT RISK 227

– Selecting funds using f.fn()

IXQGV IIQJHW 06$8;&ORVH0,23;&ORVH0**3;&ORVH0)$3;&ORVH
63<&ORVH VWDUW  HQG 
IXQGVWDLO 

msauxclose miopxclose mggpxclose mfapxclose spyclose

Date
2021-01-25 34.349998 42.810001 43.990002 26.910000 384.390015
2021-01-26 33.910000 42.509998 43.630001 26.969999 383.790009
2021-01-27 33.110001 41.509998 42.290001 26.430000 374.410004
2021-01-28 33.150002 41.810001 43.060001 26.610001 377.630005
2021-01-29 32.919998 41.250000 42.509998 26.170000 370.070007

– Creating an SMA for the Morgan Stanley Institutional Fund, Inc.

Asia Opportunity Portfolio Class A (MSAUX) using talib with
20 days
IXQGV> 0$PVDX[ @ WDOLE60$ IXQGV> PVDX[FORVH @

– Plotting the SMA for comparison (Fig. 3)

Fig. 3 MSAUX SMA

228 M. GARITA

IXQGV> PVDX[FORVH @SORW ODEHO 06$8;FORVLQJSULFH ILJVL]H 

IXQGV> 0$PVDX[ @SORW ODEHO  60$06$8;

B SOW[ODEHO 'DWH

B SOW\ODEHO 60$DQG&ORVLQJ3ULFH

B SOWWLWOH 06$8;60$DQG&ORVLQJ3ULFH

SOWOHJHQG 

– Creating an EMA for the Morgan Stanley Institutional Fund, Inc.

Asia Opportunity Portfolio Class A (MSAUX) using talib with
20 days (Fig. 4)
IXQGV> (0$PVDX[ @ WDOLE(0$ IXQGV> PVDX[FORVH @WLPHSHULRG 

IXQGV> PVDX[FORVH @SORW ODEHO 06$8;FORVLQJSULFH ILJVL]H 

IXQGV> (0$PVDX[ @SORW ODEHO  (0$06$8;

B SOW[ODEHO 'DWH

B SOW\ODEHO (0$DQG&ORVLQJ3ULFH

B SOWWLWOH 06$8;(0$DQG&ORVLQJ3ULFH

SOWOHJHQG 

Fig. 4 MSAUX EMA

 VALUE AT RISK 229

– Creating Bollinger Bands for the Morgan Stanley Institutional

Fund, Inc. Asia Opportunity Portfolio Class A (MSAUX) using
talib with 20 days (Fig. 5)
funds['up_band'], funds['mid_band'], funds['low_band'] =
talib.BBANDS(funds['msauxclose'], timeperiod =20)
funds.tail()

funds['msauxclose'].plot(label='MSAUX closing price',figsize=(16,8))

funds['up_band'].plot(label= 'Upper Band')
funds['mid_band'].plot(label= 'Midle Band')
funds['low_band'].plot(label= 'Lower Band')

_ = plt.xlabel('Date')

_ = plt.ylabel('Bollinger Bands and Closing Price')

_ = plt.title('MSAUX Bollinger Bands and Closing Price')

plt.legend();

– Creating RSI for the Morgan Stanley Institutional Fund, Inc. Asia
Opportunity Portfolio Class A (MSAUX) using talib with 14 days
(Fig. 6)

Fig. 5 MSAUX Bollinger Bands

230 M. GARITA

IXQGV> 56, @ WDOLE56, IXQGV> PVDX[FORVH @

IXQGV> 56, @SORW ODEHO 06$8;FORVLQJSULFH ILJVL]H 

B SOW[ODEHO 'DWH

B SOW\ODEHO 06$8;56,

B SOWWLWOH 06$8;56,

SOWOHJHQG 

– Calculate logarithmic returns for the funds (excluding the above

calculations)

returns = funds.to_log_returns().dropna()
returns.head()

msauxclose miopxclose mggpxclose mfapxclose spyclose

Date
2018-01-03 0.004619 0.004873 0.009265 0.002963 0.006305
2018-01-04 0.008032 0.005289 0.005693 0.006488 0.004206
2018-01-05 0.011929 0.007445 0.009994 0.005277 0.006642
2018-01-08 0.004507 −0.001310 0.002591 −0.002928 0.001827
2018-01-09 0.010067 0.005663 0.004732 0.005265 0.002261

Fig. 6 MSAUX RSI

 VALUE AT RISK 231

– Calculate a correlation matrix

UHWXUQVFRUU DVBIRUPDW I

msauxclose miopxclose mggpxclose mfapxclose spyclose

msauxclose 1.00 0.88 0.80 0.79 0.69
miopxclose 0.88 1.00 0.94 0.95 0.83
mggpxclose 0.80 0.94 1.00 0.90 0.89
mfapxclose 0.79 0.95 0.90 1.00 0.87
spyclose 0.69 0.83 0.89 0.87 1.00

– Rebasing the funds for comparison on one another (Fig. 7)

XQGVUHEDVH SORW ILJVL]H 

B SOW[ODEHO 'DWH

B SOW\ODEHO &ORVLQJ3ULFH

B SOWWLWOH 5HEDVHRIWKHFORVLQJSULFH

SOWOHJHQG 

– Calculate performance using f.fn() (Fig. 8)

IXQGV IXQGVGURSQD

SHUIRUPDQFH IXQGVFDOFBVWDWV

SHUIRUPDQFHSORW ILJVL]H 

B SOW[ODEHO 'DWH

B SOWWLWOH )XQGVPRQWO\3URJUHVVLRQ

SOWOHJHQG 

– Display performance indicators

performance.display()
232 M. GARITA

Fig. 7 Rebase of the closing price in funds and ETF

Fig. 8 Funds and ETF montlhy progression

Stat msauxclose miopxclose mggpxclose mfapxclose spyclose

Start 2018-01-02 2018-01-02 2018-01-02 2018-01-02 2018-01-02
End 2021-01-29 2021-01-29 2021-01-29 2021-01-29 2021-01-29
Risk-free rate 0.00% 0.00% 0.00% 0.00% 0.00%
Total Return 90.51% 83.17% 88.43% 55.31% 37.69%
Daily Sharpe 1.05 1.04 1.02 0.85 0.57
Daily Sortino 1.74 1.61 1.58 1.28 0.84
CAGR 23.32% 21.76% 22.88% 15.40% 10.96%
Max Drawdown −31.35% −28.34% −27.19% −27.78% −34.10%
Calmar Ratio 0.74 0.77 0.84 0.55 0.32
 VALUE AT RISK 233

Stat msauxclose miopxclose mggpxclose mfapxclose spyclose

MTD 5.21% 1.68% −1.23% −1.54% −1.02%
3m 15.31% 15.77% 9.08% 11.27% 12.15%
6m 27.60% 26.57% 21.81% 17.72% 13.83%
YTD 5.21% 1.68% −1.23% −1.54% −1.02%
1Y 58.42% 58.96% 47.04% 29.55% 13.30%
3Y (ann.) 21.97% 20.13% 20.80% 14.13% 9.14%
5Y (ann.) 23.32% 21.76% 22.88% 15.40% 10.96%
10Y (ann.) 23.32% 21.76% 22.88% 15.40% 10.96%
Since Incep. 23.32% 21.76% 22.88% 15.40% 10.96%
(ann.)
Daily Sharpe 1.05 1.04 1.02 0.85 0.57
Daily Sortino 1.74 1.61 1.58 1.28 0.84
Daily Mean (ann.) 23.51% 21.93% 23.22% 16.13% 13.03%
Daily Vol (ann.) 22.41% 20.98% 22.68% 18.87% 22.79%
Daily Skew −0.34 −0.76 −0.61 −1.00 −0.70
Daily Kurt 2.10 9.41 8.60 14.99 13.18
Best Day 5.95% 8.13% 7.83% 7.37% 9.06%
Worst Day −7.92% −9.93% −10.55% −9.81% −10.94%
Monthly Sharpe 1.06 1.05 1.08 0.95 0.58
Monthly Sortino 2.10 2.14 2.27 1.72 0.98
Monthly Mean 21.90% 20.24% 20.82% 14.31% 10.82%
(ann.)
Monthly Vol 20.62% 19.35% 19.35% 15.05% 18.72%
(ann.)
Monthly Skew −0.49 −0.21 −0.04 −0.58 −0.38
Monthly Kurt −0.29 −0.15 −0.02 0.02 0.61
Best Month 10.70% 11.89% 12.75% 7.81% 12.70%
Worst Month −13.87% −11.99% −11.10% −9.43% −13.00%
Yearly Sharpe 1.36 1.13 1.05 1.07 0.98
Yearly Sortino inf inf 40.77 22.41 24.89
Yearly Mean 33.30% 30.38% 28.99% 19.96% 14.64%
Yearly Vol 24.53% 26.77% 27.70% 18.65% 14.96%
Yearly Skew −1.60 −0.72 −0.93 −1.70 −0.45
Yearly Kurt - - - - -
Best Year 50.51% 54.67% 53.17% 31.85% 28.79%
Worst Year 5.21% 1.68% −1.23% −1.54% −1.02%
Avg. Drawdown −3.55% −3.14% −3.38% −2.30% −2.64%
Avg. Drawdown 30.15 26.30 22.62 19.40 22.34
Days
Avg. Up Month 5.79% 5.39% 5.63% 3.81% 3.84%
Avg. Down −4.41% −3.49% −3.13% −3.44% −4.97%
Month
Win Year % 100.00% 100.00% 66.67% 66.67% 66.67%
Win 12 m % 80.77% 76.92% 84.62% 80.77% 88.46%

– Calculate the drawdown series

funds.to_drawdown_series().tail()
234 M. GARITA

msauxclose miopxclose mggpxclose mfapxclose spyclose

Date
2021-01-25 0.000000 0.000000 −0.003398 −0.006645 0.000000
2021-01-26 −0.012809 −0.007008 −0.011554 −0.004430 −0.001561
2021-01-27 −0.036099 −0.030367 −0.041912 −0.024363 −0.025963
2021-01-28 −0.034934 −0.023359 −0.024468 −0.017719 −0.017586
2021-01-29 −0.041630 −0.036440 −0.036928 −0.033961 −0.037254

The example above that compared funds and an ETF, compared in

performance the MSAUX as the best option based on the total return,
the CAGR and the 12-month percentage. The data also shows that it
is the riskier asset in its class with an average drawdown of −3.55%, the
highest average drawdown days with 30.15 and a high average draw-
down per month. The information of the performance report can be
applied to everything learned in this book.

Works Cited
Damodaran, Aswath. 2018. Value at Risk (VAR). New York University. n.d.
Accessed February 20, 2019. http://people.stern.nyu.edu/adamodar/
pdfiles/papers/VAR.pdf.
Mitchell, Cory. 2019. Drawdown definition and example. 25 June. Accessed July
30, 2019. https://www.investopedia.com/terms/d/drawdown.asp.
 Works Cited

365 Data Science. 2020. Why Python for data science and why Jupyter to code in
Python Articles 11 min read. n.d. Accessed March 2, 2020. https://365datasci-
ence.com/why-python-for-data-science-and-why-jupyter-to-code-in-python/.
Anaconda. 2020. Anaconda distribution. Accessed March 2, 2020. https://
www.anaconda.com/distribution/.
Bang, Julie. 2019. Candlestick bar. Investopedia.
Basurto, Stefano. 2020. Python trading toolbox: Introducing OHLC charts with
Matplotlib. 07 January. Accessed March 31, 2020. https://towardsdatasci-
ence.com/trading-toolbox-03-ohlc-charts-95b48bb9d748.
Bloomberg Corporation. 2019. Bloomberg puts the power of Python XE “Python”
in hedgers’ hands. 8 March. Accessed March 2, 2020. https://www.bloomb-
erg.com/professional/blog/bloomberg-puts-power-python-hedgers-hands/.
Bollinger, John. 2018. John Bollinger answers “What are Bollinger Bands?”. n.d.
Accessed March 2, 2020. https://www.bollingerbands.com/bollinger-bands.
Bolsa de Madrid. 2020. Electronic Spanish Stock Market Interconnection System
(SIBE). n.d. Accessed March 23, 2020. http://www.bolsamadrid.es/ing/
Inversores/Agenda/HorarioMercado.aspx.
Brooks, Chris. 2008. Introductory econometrics for finance. Boston: Cambridge
University Press.
Bryant, Bradley James. 2020. How to calculate portfolio value. n.d. Accessed January
3, 2020. https://www.sapling.com/5872650/calculate-portfolio-value.
Burgess, Matthew, and Sarah Wells. 2020. Giant wealth fund seeks managers who
can beat frothy market. 9 February. Accessed March 2, 2020. https://finance.
yahoo.com/news/giant-wealth-fund-seeks-managers-230000386.html.
Chen, James. 2019a. Financial risk. 15 June. Accessed August 20, 2019.
https://www.investopedia.com/terms/f/financialrisk.asp.

© The Editor(s) (if applicable) and The Author(s) 2021 235

M. Garita, Applied Quantitative Finance,
https://doi.org/10.1007/978-3-030-29141-9
236 Works Cited

Chen, James. 2019b. Jensen’s Measure. 21 November. Accessed December 20,

2019. https://www.investopedia.com/terms/j/jensensmeasure.asp.
Chen, James. 2019c. Line Chart. 12 August. Accessed March 25, 2020.
https://www.investopedia.com/terms/l/linechart.asp.
Damodaran, Aswath. 2018. Value at Risk (VAR). New York University. n.d.
Accessed February 20, 2019. http://people.stern.nyu.edu/adamodar/
pdfiles/papers/VAR.pdf.
Das, Sejuti. 2020. Analytics India Magazine. 27 02. Accessed March 17, 2020.
https://analyticsindiamag.com/why-jupyter-notebooks-are-so-popular-
among-data-scientists/.
Mullins, David W. 1982. Does the capital asset pricing model work? n.d January.
Accessed August 13, 2019. https://hbr.org/1982/01/does-the-capital-
asset-pricing-model-work.
Fontinelle, Amy. 2019. Systematic risk. 30 September. Accessed October 1, 2019.
https://www.investopedia.com/terms/s/systematicrisk.asp.
Ganti, Akhilesh. 2019. Central Limit Theorem (CLT). 13 September. Accessed
April 2, 2019. https://www.investopedia.com/terms/c/central_limit_theo-
rem.asp.
Halton, Clay. 2019. Line graph. 21 August. Accessed March 25, 2020. https://
www.investopedia.com/terms/l/line-graph.asp.
Hargrave, Marshall. 2019. Sharpe ratio. 17 March. Accessed April 1, 2019.
https://www.investopedia.com/terms/s/sharperatio.asp.
Hargrave, Marshall. 2020. Standard deviation definition. 1 February. Accessed
February 20, 2020. https://www.investopedia.com/terms/s/standarddevia-
tion.asp.
Hayes, Adam. 2018. Volume definition. 4 February. Accessed March 25, 2020.
https://www.investopedia.com/terms/v/volume.asp.
Hayes, Adam. 2019a. Correlation definition. 20 June. Accessed January 1, 2020.
https://www.investopedia.com/terms/c/correlation.asp.
Hayes, Adam. 2019b. Correlation definition. 20 June. Accessed October 8,
2019. https://www.investopedia.com/terms/c/correlation.asp.
Hayes, Adam. 2019c. Variance. 2 September. Accessed August 3, 2019. https://
www.investopedia.com/terms/v/variance.asp.
Hayes, Adam. 2020a. Exponential Moving Average - EMA definition. 8 July.
Accessed April 2, 2020. https://www.investopedia.com/terms/e/ema.asp.
Hayes, Adam. 2020b. Moving Average (MA). 31 March. Accessed March 31,
2020. https://www.investopedia.com/terms/m/movingaverage.asp.
Hunner, Trey. 2018. Trey Hunner. 11 October. Accessed March 23, 2020.
https://treyhunner.com/2018/10/asterisks-in-python-what-they-are-
and-how-to-use-them/.
 Works Cited 237

Jain, Diva. 2018. Skew and Kurtosis: 2 Important statistics terms you need to know
in Data Science. 23 August. Accessed August 12, 2019. https://codeburst.
io/2-important-statistics-terms-you-need-to-know-in-data-science-skewness-
and-kurtosis-388fef94eeaa.
Kalla, Siddharth. 2020. Range (Statistics). n.d. Accessed January 4, 2020.
https://explorable.com/range-in-statistics.
Kan, Chi Nok. 2018. Data Science 101: Is Python better than R? 1 August.
Accessed March 2, 2020. https://towardsdatascience.com/data-science-101-
is-python-better-than-r-b8f258f57b0f.
Keaton, Will. 2019. Quantitative Analysis (QA). 18 April. Accessed January 15,
2020. https://www.investopedia.com/terms/q/quantitativeanalysis.asp.
Keaton, Will. 2020. Treynor ratio. 22 March. Accessed Apri 1, 2020. https://
www.investopedia.com/terms/t/treynorratio.asp.
Kenton, Will. 2019. Kurtosis. 17 February. Accessed July 30, 2019. https://
www.investopedia.com/terms/k/kurtosis.asp.
Kevin, S. 2015. Security analysis and portfolio management. Delhi: PHI.
London Stock Exchange Group. 2020. London Stock Exchange Group Business
Day. n.d. Accessed March 25, 2020. https://www.lseg.com/areas-expertise/
our-markets/london-stock-exchange/equities-markets/trading-services/
business-days.
Mastromatteo, Davide. 2020. Python args and kwargs: Demystified. 09 September.
Accessed March 31, 2020. https://realpython.com/python-kwargs-and-args/.
Milton, Adam. 2020. Simple, exponential, and weighted moving averages. 09
November. Accessed March 31, 2020. https://www.thebalance.com/
simple-exponential-and-weighted-moving-averages-1031196.
Mitchell, Cory. 2019a. Don’t trade based on MACD divergence until you read
this. 19 November. Accessed April 1, 2020. https://www.thebalance.com/
dont-trade-based-on-macd-divergence-until-you-read-this-1031217.
Mitchell, Cory. 2019b. Drawdown definition and example. 25 June. Accessed
July 30, 2019. https://www.investopedia.com/terms/d/drawdown.asp.
Mitchell, Cory. 2019. Understanding basic candlestick charts. 19 December.
Accessed March 30, 2020. https://www.investopedia.com/trading/
candlestick-charting-what-is-it/.
Mitchell, Cory. 2020. How to use volume to improve your trading. 25 February.
Accessed March 27, 2020. https://www.investopedia.com/articles/techni-
cal/02/010702.asp.
Mitra, Gautam, and Leela Mitra. 2011. The handbook of news analytics in finance.
United Kingdom: Wiley.
Murphy, Casey. 2020. Investopedia. 16 November. Accessed January 01, 2021.
https://www.investopedia.com/trading/introduction-to-parabolic-sar/.
238 Works Cited

Murphy, Chris. 2019. Information Ratio – IR. 10 January. Accessed February

20, 2019. https://www.investopedia.com/terms/i/informationratio.asp.
Murphy, John J. 1999. Technical analysis of the financial markets. New York:
New York Institute of Finance.
NYSE. 2020. TAQ closing prices. n.d. Accessed March 25, 2020. https://www.
nyse.com/market-data/historical/taq-nyse-closing-prices.
O`Reilly. 2020. 5 key areas for tech leaders to watch in 2020. 18 February. Accessed
March 2, 2020. https://www.oreilly.com/radar/oreilly-2020-platform-analysis/.
Pfeiffer, Frank. 2019. R versus Python: Which programming language is better for
data science projects in Finance? 28 May. Accessed March 2, 2020. https://
finance-blog.arvato.com/r-versus-python-in-finance/.
Posey, Luke. 2019. Implementing MACD in Python XE “Python”. 30
March. Accessed April 2, 2020. https://towardsdatascience.com/
implementing-macd-in-python-cc9b2280126a.
Python Organization. 2020. Python organization. 07 January. Accessed March
02, 2020. https://docs.python.org/2/faq/general.html#what-is-python.
Python. 2006. 2.3.4 Numeric types -- int, float, long, complex. 18 October.
Accessed March 23, 2020. https://docs.python.org/2.4/lib/typesnumeric.
html.
Python. 2020. Data structures. 23 March. Accessed March 23, 2020. https://
docs.python.org/3/tutorial/datastructures.html.
Shaik, Naushad. 2018. 5 reasons why learning Python is the best decision. 21
September. Accessed March 2, 2020. https://medium.com/datadrivenin-
vestor/5-reasons-why-i-learned-python-and-why-you-should-learn-it-as-well-
917f781aea05.
Trochim, William M.K. 2020. Correlation. 10 March. Accessed March 12, 2020.
https://conjointly.com/kb/correlation-statistic/.
Wan, Xiang, Wengian Wang, Jiming Liu, and Tiejun Tong. 2014.
Estimating the sample mean and standard deviation from the sample
size, median, range and/or interquartile range. 19 December. Accessed
January 03, 2019. https://bmcmedresmethodol.biomedcentral.com/
articles/10.1186/1471-2288-14-135.
 Index

A datetime, 52, 62, 76, 86, 87, 102,

addition, 20, 22 103, 105, 108, 110, 120, 123,
Anaconda, 81 124, 130, 135, 136, 140, 141,
api, 39, 59, 60, 92, 95, 113, 146, 177 143, 147, 149, 158, 160, 163,
append, 32, 56, 97, 98 166, 172, 178, 180, 183, 184,
array, 25, 37, 40, 174, 184, 190, 195, 186, 189, 191, 195, 198, 199,
198, 202, 212, 215, 219 202, 205, 212, 214, 218, 219,
222, 223, 226
dictionary, 33–37, 51, 138
B dividing, 20, 22
Boolean, 39, 41, 42, 45 Dow Jones, 27–29, 137, 141, 162
drop, 38, 41, 208, 213

C
Central Limit Theorem, 85 E
complex, 20, 21, 23, 25, 34, 216 Elif, 45
Else, 44
Excel, ix, 2, 6, 7, 71, 81, 83, 119
D
DataFrame, 23, 24, 26, 27, 31,
36–39, 52, 64, 67, 69, 79, 82, F
92, 95, 110, 111, 144, 146, f.fn, 36, 73, 74, 79, 80, 88, 91, 93, 96,
149, 154, 172, 177, 178, 112, 116, 222, 223, 226, 227, 231
180, 183, 184, 189, 195, float, 20, 21, 23, 25, 66, 67
198, 202, 212, 214, 219, for loop, 46–48, 50, 52, 55, 177, 178,
223, 225 180, 189

© The Editor(s) (if applicable) and The Author(s) 2021 239

M. Garita, Applied Quantitative Finance,
https://doi.org/10.1007/978-3-030-29141-9
240 Index

FRED, 59–63, 68 P
fredapi, 60, 62 Pandas, 73, 75, 89, 90, 92, 95, 102,
107, 149, 154
pandas_datareader, 62, 76, 86, 102,
G 105, 108, 110, 120, 123, 129,
GDP deflator, 65 135, 136, 140, 143, 147, 149,
Google Colab, 14, 16, 73, 79 158, 160, 163, 179, 182, 189,
The Gross Domestic Product, 63, 65 212, 214, 218, 222, 226
PyNance, 74
Python, vii, x, 1–7, 8–10, 11, 14,
H 19–22, 25, 27, 30, 32, 33, 35,
histogram, 87–89, 91–94, 96–99, 44–48, 51, 62, 66, 71–75, 77,
106, 107, 113, 157 81, 83, 85, 87, 92, 95, 100, 101,
119, 121, 154, 162, 172, 177

I
Indexing, 27, 28 Q
integer, 20–23, 31, 66, 67 QuantPy, 74

L R
len, 26–28 returns, 89
list, 25–33, 36, 47, 50, 51, 55–57, 65,
67, 68
List Comprehension, 55 S
Loops, 46 S&P 500, 28, 29, 110, 173, 174, 177,
181, 186, 188, 191, 195, 199, 202
SciPy, 74
M square root, 20, 22
Matplotlib, 4, 73–75 Sturge’s Rule, 89–91
mean, 37, 40, 79, 86, 87, 94–96, 105, subtracting, 20, 22
106, 144, 149, 154, 192, 193,
197, 200, 203, 204, 208, 215,
216, 223 T
median, 87 Ta-lib, 73
mode, 87 TIA, 74
multiply, 37, 40, 56, 128, 173 tickers, 52, 172, 178, 180, 183, 184,
multiplying, 20, 22 189, 195, 198, 202, 212, 214,
219, 223

N
Natural Logarithm, 92 V
NumPy, 11, 72, 74, 75 Value At Risk, 42