PREFACE

A lgorithmic Trading (Algo or High Frequency Trading) is a technology platform providing an

advantage of both the worlds – Artificial Intelligence and Human Intelligence. Algorithms
are Mathematics and Rules; and Rules and Logic, Human Programming is responsible for the
trade transactions. Here, the complex mathematical models and formulas enables the trader to make
high-speed decisions and transactions in the financial markets, Given the advantages of such
technological interventions to facilitate securities transactions in the market, Algo or High Frequency
Trading has of late become the buzz word in the trading ecosystem. In developed countries, the volume
of such technology-driven transactions is significantly increasing over the years. Indian market is also
gradually embracing the Algo or High Frequency Trading, here also more and more transactions are
being routed through this platform. Indeed, there are considerable advantages of these trading practices,
yet at the same time, there are areas of serious concerns as well, and they may lead to market wide
systemic risks besides others.

Every information has an underlying cost, Algo or High Frequency Trading software technology being
still in the evolving stage, the obsolesce rate is not only very high besides generally being event-based
software, the acquisition cost of the technology is also on the higher side. As a result, the institutional
investors are the beneficiaries of this technology. Retail investors continued to be deprived of
technological advancements in the trading models. But then, one must concede that this trading
technology is going to stay.

Recent market developments in India have heightened the concerns of the Policy Makers and Market
Regulators is a trading practice which is vulnerable enough to need regulatory protection.

This Report comprehensive deliberates on this contemporary market trading practices, benefits and
areas of concern. While drawing conclusions, the Report makes an endeavour to propose improvements
in Policy Framework for Algo or High Frequency Trading for the benefit of the Policy Makers.

DEA-NIFM Research Team

i
 ACKNOWLEDGMENT

N
ational Institute of Financial Management is pleased to be provided with an opportunity by the
Department of Economic Affairs to make a purposeful academic contribution to one of the most
contemporary issues of the Securities Markets – Algorithm Trading/High Frequency Trading.

Technology has made Artificial Intelligence as all-pervasive market wide. Developed markets and
emerging markets have embraced this technology in the securities market. Market innovations and
complexities have ensured that Algo Trading is fact-of-the -day and going to stay. As the cliché goes,
every Coin has two sides. Be that so, the best insulation is that such trading must be Fair, Transparent,
Accountable and above all ensure Investors’ Protection. In the domestic market context, certain market
developments have made Algo Trading and its fall out as an area of immense academic research and
generated scope of future-learnings.

Being an area where a restricted research has been conducted and debated, the Research Team had to
mine information/data from various global resources, which are generally protected. Understanding and
appreciating global developments and juxtaposing them with the Indian securities market ecosystem
involved not only super-specialised domain expertise but also out-of-box thinking with a forward-
looking approach. Given that the technology obsolescence rate is very high, the team had to anticipate
the prospective regulatory framework for the Indian market. Hurdles were many on the way,
perseverance paved the way to the successful culmination of the assignment.

We place on record our deep and sincere appreciation to the entire team of National Stock Exchange
and Bombay Stock Exchange for providing us access to the inside of the regulatory platform and also
to share all the requisite information/details that were necessary in writing this report.

Our report would not have been made possible but for the unprecedented contributions of Mr. Harjeet
Singh, Mr. Kunal Nandwani, Consultants, DEA-NIFM Research Programme and also experts from the
Fintech sector of the Capital Markets in preparing this Report and providing significant inputs in
drafting the proposed Regulatory Framework for the Policy Makers and the Market Regulators alike
for making the domestic markets more attractive and investor friendly.

Prof. (Dr.) A M Sherry

Chair, DEA-NIFM Research Programme

ii
 EXECUTIVE SUMMERY

WHAT IS ALGORITHMIC TRADING?

Algorithmic trading is the use of programs and computers to generate and execute (large) orders in
markets with electronic access. Orders come from institutional investors, funds and trading desks of big
banks and brokers. These statistical, mathematical or technical models analyze every quote and trade in
the stock market, identify liquidity opportunities, and turn the information into intelligent trading
decisions.

Algorithmic trading, or computer-directed trading, cuts down transaction costs, and allows investment
managers to take control of their own trading processes. The main objective of algo trading is not
necessarily to maximize profits but rather to control execution costs and market risk.

ALGORITHMIC TRADING AND ITS COMPOSITION IN INDIAN MARKETS

Around 50% plus of total orders at both NSE and BSE are algo trades on the client side. Prop side algo
trades are 40% plus of total orders placed at both the exchanges. More than 80% of the algorithmic
orders are generated from colocation at both the exchanges. In developed markets it stands at about
80%.

KINDS OF ALGORITHMS

Algorithms are used extensively in various stages of the trading cycle. We can classify them into pre-
trade analytics, execution stage, and post-trade analytics.

Depending on their usage, Algorithms can also be broadly classified into Agency trading algorithms,
Proprietary Trading algorithms and High Frequency Trading (HFT) algorithms.

Execution Algorithms - Execution algorithms mean to systematically split a larger order into many
smaller orders based on the available liquidity. These amounts are often larger than what the market
can absorb without impacting the price. For instance, Time Weighted Average Price (TWAP)
algorithmic strategy will break an order up into many smaller equal parts and execute them during the
trading day, normally at 5 minute intervals. Another example is of the Volume Weighted Average Price
(VWAP) strategy that will estimate the average volume traded for each 5-minute interval the order is
traded using historical trading information, with the ultimate goal to split the order into smaller pieces
based on an average weighted volume.

Proprietary Trading Algorithms - Proprietary trading (also "prop trading") occurs when a trader
trades stocks, bonds, currencies, commodities, their derivatives, or other financial instruments with the
firm's own money, as opposed to depositors' money, so as to make a profit for itself. Proprietary Trading
algorithms are typically used with the strategies that involve directional bets on the markets – Net Long
or Short depending on the market direction. Within this subset, we have Momentum, Mean Reversion
and Trend Following strategies. Besides, another popular set of strategies called as Spread strategies or
Market Neutral (both Long/Short) is also part of this suit of algorithms.

HFT Algorithms - High-frequency trading (HFT) is a subset of automated trading. Here,

opportunities are sought and taken advantage of on very small timescales from nanoseconds up to

iii
milliseconds. Some high-frequency strategies adopt a market maker type role, attempting to keep a
relatively neutral position and proving liquidity (most of the time) while taking advantage of any price
discrepancies. Other strategies invoke methods from time series analysis, machine learning
and artificial intelligence to predict movements and isolate trends among the masses of data.

HOW IS AN ALGORITHM BUILT?

Decide upon the genre/strategy paradigm - The first step is to decide the strategy paradigm. It can be
Market Making, Arbitrage based, Alpha generating, Hedging or Execution based strategy.

Establish Statistical significance - You can decide on the actual securities you want to trade based on
market view. Establish if the strategy is statistically significant for the selected securities

Build Trading model – Next step would be to code the logic based on which you want to generate
buy/sell signals in your strategy.

Quoting or Hitting strategy - It is very important to decide if the strategy will be “quoting” or “hitting”.
Execution strategy to a great extent decides how aggressive or passive your strategy is going to be.

Backtesting & Optimization – This step is extremely important to understand if the strategy you chose
works well in the markets or not. A strategy can be considered to be good if the backtest results and
performance statistics back the hypothesis.

COLOCATION AND IMPLICATIONS

Colocation is locating computers owned by HFT firms and proprietary traders in the same premises
where an exchange’s computer servers are housed. This enables HFT firms to access stock prices a split
second before the rest of the investing public.

Co-location has become a lucrative business for exchanges, which charge HFT firms by rack space for
the privilege of “low latency access.”

Colocation reduces latency, increases liquidity and levels the playing field among competing HFT
market makers.

In the Indian context, the disadvantages of colocation include expensive and market inequality.

ADVANTAGES AND DISADVANTAGES OF ALGORITHMIC TRADING

Algo trading, colocation and HFT offer various advantages and disadvantages. It is observed that with
algo trading and HFT there have been improvements in transactions costs, volatility, and buy-sell
imbalance. Market prices have become more efficient and they have facilitated price discovery.
Algorithms using Colocation reduce latency and enhance liquidity.

Lack of control has led to systemic risks. Fat finger or faulty algorithms can cause huge deviations from
healthy prices. Examples include Flash crash that occurred on BSE Muhurat Session in 2011, Flash
crash on Nifty April futures on April 21st, 2012 and Reliance Industries stock flash crash on June 2010,
due to execution of a large ‘sell’ order using algorithms.

iv
It has been proved in the past that Algo tading and HFT can be used to manipulate markets using
techniques like quote stuffing, layering (spoofing) and momentum ignition. Evidence suggests that
market manipulation algorithms lead to decreased liquidity, higher trading costs, increased short term
volatility, impact performance and fill rates, and massive price moves backed by false volume.

ORDER TO TRADE RATIO AND SIGNIFICANCE

Order-to-trade (or order-to-execution) ratios involve financially penalising individual financial firms if
the orders to buy or sell they enter do not lead to a ‘sufficient’ number of trades.

High order-to-trade ratios imply that market participants are placing and cancelling orders but not
executing most of the orders. This could be due to the nature of market making algorithms or market
manipulation algorithms, where orders are placed to drive volumes to that point and then cancelled –
with the result that most of the orders are not converted into trades.

In the year 2016-17, order to trade ratio for NSE across all segments was 11.2. It has increased from
7.07 in 2014-15. In case of high order to trade ratio, NSE makes calls and alert trading members. BSE
has issued circulars to keep a check on high order to trade ratios. Penalty is imposed by both the
exchanges for high to trade ratios for the member brokers.

MEASURES ADOPTED BY SECURITIES MARKET REGULATORS IN DIFFERENT

COUNTRIES

Minimum resting time, frequent batch auctions, random speed bumps or delays, randomization of orders
during a period (1-2 seconds), max order message to trade ratio requirement, market maker pricing are
some of the measures adopted globally by various market regulators.

Some of the other important measures carried out include the HFT transaction tax implemented by the
regulators in France and Italy; Market Access Rule, Regulation SCI and registration of entities
implemented by SEC; and the enactment of the Act on the Prevention of Risks and Abuse in High-
Frequency Trading (HFT Act) by the German regulators in 2013.

SURVEILLANCE METHODS AT NSE & BSE

Currently both NSE and BSE have their own methods and levels of sophistication to manage
surveillance. However, in our view, harmonization of surveillance mechanism would bring about
uniformity in exchange action towards harmful HFT. There is a definite need to invest in advanced
technology to automatically detect harmful HFT and market manipulative trends/algorithms.
Exchanges hardly have advanced real-time surveillance mechanisms to detect harmful HFT.

SEBI’S DISCUSSION PAPER ON CURBING HARMFUL HFT

Minimum Resting Time: Resting time is defined as the time between an order is received by the
exchange and the said order is allowed to be amended or cancelled thereafter. Securities market
regulators have considered the idea of eliminating “fleeting orders” or orders that appear and then
disappear within a short period of time. As per the Minimum Resting Time mechanism, the orders
received by the stock exchange would not be allowed to be amended or cancelled before a specified
amount of time viz. 500 milliseconds is elapsed. Currently, there are no instances of the ‘resting time’
mechanism being mandated by any regulator. It has been observed that Australian Securities and
Investment Commission (ASIC) had sought feedback on the matter few years ago, but decided not to
go ahead with the proposal. Perceived advantages include more stability in limit order book, reduce

v
fleeting orders, and reduce the excessive level of message traffic. Perceived disadvantages include
longer queues & waiting time, rising transaction costs and increased volatility.

Frequent Batch Auctions: The mechanism of Frequent Batch Auctions would accumulate buy and
sell orders on the order book for a particular length of time (say 100 milliseconds). At the end of every
such period, the exchange would match orders received during the time interval. This proposal tries to
address the problem of ‘latency advantage’ by undertaking batch auctions at a particular interval. The
idea is to set a time interval for matching of orders which is short enough to allow for opportunities for
intraday price discovery, but long enough to minimize the latency advantage of HFT to a large extent.
There is no evidence of implementation of Frequent Batch auctions. Perceived advantages include
reduction of the speed of trading and elimination of sniping. Perceived disadvantages include
impediment of price discovery, increased execution risk and reduced liquidity.

Random Speed Bumps: The Speed Bump mechanism involves introduction of randomized order
processing delay of few milliseconds to orders. Instances where this mechanism has been implemented
include TSXA – Toronto Stock Exchange (1-3 ms) and ParFX – interdealer OTC broker (20-80ms)
impose randomized order processing. Perceived advantages include nullify latency advantage, market
equity and stop arms race for speed. Perceived disadvantages include reduction or withdrawal of
liquidity.

Randomization of orders received during a period (1-2seconds): The time-priority of the new /
modified orders that would be received during predefined time period (say 1-2 seconds period) is
randomized and the revised queue with a new time priority is then forwarded to the order matching
engine. Instances where this mechanism has been implemented include ICAP EB (wholesale FX
electronic trading platform) Market Matching platform that has introduced the Latency floor. Perceived
advantages include reduce latency advantage, market equity and stop arms race for speed. Perceived
disadvantages include reduction or withdrawal of liquidity.

Maximum order to trade ratio requirement: A maximum order-to-trade ratio requires a market
participant to execute at least one trade for a set number of order messages sent to a trading venue. The
mechanism is expected to increase the likelihood of a viewed quote being available to trade and reduce
hyper-active order book participation. NSE and BSE are already implementing this mechanism –
disincentives include penalty charges and trading ban for 15 minutes on the subsequent trading session.
Perceived advantages include increased market depth, curtailed market manipulation and reduced large
number of electronic messages. Perceived disadvantages include reduction in liquidity, withdrawal
during volatile times and increased bid-ask spread.

Separate queues for colo and non-co-location orders: With the view to ensure that stock brokers who
are not co-located have fair and equitable access to the stock exchange’s trading systems, stock
exchanges facilitating co-location shall implement an order handling architecture comprising of two
separate queues for co-located and non-colocated orders such that orders are picked up from each queue
alternatively. It is expected that such architecture will provide orders generated from a non-colocated
space a fair chance of execution and address concerns related to being crowded-out by orders placed
from colocation. Perceived advantages include provide fair chance for non-Co-location orders.
Perceived disadvantages include creation of 2 parallel markets and possible withdrawal of liquidity.

Review of tick by tick data feed: Tick-by-Tick (TBT) data feed provide details relating to orders
(addition + modification + cancellation) and trades on a real-time basis. TBT data feed facilitates a
detailed view of the order-book (such as depth at each price point, etc.). TBT facility is being provided

vi
by BSE and NSE to collocated participants. Perceived advantages include more transparency, access to
full order book and real time access to TBT data. Perceived disadvantages include reduce the level of
transparency if the data feed is anything other than real time feed and the increased analytical capacity
required at the brokers end to sift through TBT data.

vii
 ABBREVIATIONS

AI Artificial Intelligence

ARC Annual Recovery Charges

BSE Bombay Stock Exchange

BEFS BSE Electronic Filing System

CM Capital Market

CAT Consolidated Audit Trail

CAGR Compound Annual Growth Rate

CDs Currency Derivatives

DMA Direct memory access

EBS Electronic Broking Services

ETPs Exchange Traded Products

ECN Electronic Communication Network

ETFs Exchange Traded Funds

F&O Futures and Options

HFT High-frequency trading

IOSCO International Organization Of Securities Commissions

IPO Initial Public Offering

IOC Immediate or Cancel

IFSD Interest Free Security Deposit

MOC Market-On-Close

MFR Monthly Fill Ratio

MQL Minimum Quote Lifespan

MPLS Multiprotocol label switching

NYSE New York Stock Exchange

NASDAQ National Association of Securities Dealers Automated Quotations

NSE National Stock Exchange

viii
NSEIL National Stock Exchange of India Ltd.

OTC Over the Counter

OER Orders-to-Executions

PoV Percent of Value

RTPC Reverse Trade Prevention Check

SEC Securities Exchange Commission

SSF Single Stock Futures

STPC Self-Trade Prevention Check

SMA Sponsored Market Access

TOR Terms of Reference

TWAP Time Weighted Average Price

TBT Tick-by-Tick

TWSE Taiwan Stock Exchange

TSXA TSX Alpha Exchange

VSAT Very Small Aperture Terminal

VaR Value-at-Risk

VWAP Volume Weighted Average Price

ix
 CONTENTS

PREFACE…………………………..……………………………………………………….…………….... i
ACKNOWLEDGEMENT ………………………………………………………………………………......ii
EXECUTIVE SUMMERY …………………………………………………………………………………iii
ABBREVIATIONS ……………………………………………………………………………………......viii
LIST OF TABLES …………………………………………………………………………………………xiv

CHATPER1: ALGORITHMS…………………………………………………………..……………..…......1

1.1 INTRODUCTION TO ALGORITHMS ............................................................................................. 2

1.2 TYPES OF ALGORITHMS ............................................................................................................... 2
1.3 ADVANTAGES OF ALGORITHMS .................................................................................................. 4
1.4 DISADVANTAGES OF ALGORITHMS ........................................................................................... 5
1.5 CLASSIFICATION OF ALGORITHMS BASED ON LIFE-CYCLE STAGE .................................. 7
1.6 LIFECYLE OF AN ALGORITHM ...................................................................................................... 9
1.7 WHAT ARE EXECUTION ALGORITHMS AND HOW ARE THEY DEVELOPED ................... 11
1.8 WHAT ARE HFT ALGORITHMS AND HOW ARE THEY DEVELOPED................................... 13

CHAPTER 2: COMPOSITION OF ALGORITHMS TRADING………………………………………..…14

2.1 COMPOSITION OF ALGO TRADING TAKING PLACE IN THE COUNTRY-CLIENT OR

PROPRIETARY ............................................................................................................................... 15
2.2 CURRENTLY REGISTERED ALGORITHM TRADING PLAYERS............................................. 15
2.3 TOP 20 ALGO PARTICIPANTS AND THEIR DAILY TURNOVER ............................................ 16
2.4 ALGORITHMIC TRADING TRENDS AND THE EXTENT TAKING PLACE IN INDIAN
MARKETS VIS-À-VIS GLOBAL MARKETS ............................................................................... 17
2.5 COMPARISON OF INDIAN MARKETS VIS-À-VIS GLOBAL MARKETS ................................ 19

CHAPTER 3: HIGH FREQUENCY TRADING (HFT)………………………………………..…..………20

3.1 INTRODUCTION TO HFT ALGORITHMS .................................................................................... 21

3.2 BENEFITS OF HFT ........................................................................................................................... 21
3.3 DEMERITS OF HFT .......................................................................................................................... 23
3.4 HFT IMPACT ON INSTITUTIONAL INVESTORS ........................................................................ 24
3.5 HFT IMPACT ON VOLATILITY ..................................................................................................... 26
3.6 MISUSING HFT FOR MARKET MANIPULATION ...................................................................... 28
3.7 MARKET MANIPULATION TECHNIQUES USING HFT ............................................................ 29

CHAPTER 4: CO-LOCATION…………………………………………………………………………….35

4.1 WHAT IS CO-LOCATION……………….…………………….......................................................36

x
 4.2 ADVANTAGES OF CO-LOCATION ............................................................................................... 36
4.3 DISADVANTAGES OF CO-LOCATION ........................................................................................ 37
4.4 CO-LOCATION FACILITIES IN INDIA ......................................................................................... 37
4.5 CO-LOCATION ARCHITECTURE .................................................................................................. 38
4.6 QUANTUM OF ORDERS EMANATING FROM CO-LOCATION ............................................... 40
4.7 CO-LOCATION COSTS AT NSE……………………………………………………………..........40

4.8 COLOCATION COSTS AT BSE ...................................................................................................... 42

4.9 REQUIREMENTS TO BE MET FOR SETTING UP COLOCATION SERVER ............................ 43
4.10 LENGTH AND QUALITY OF WIRE USED ................................................................................. 43
4.11 SYSTEM (UNICAST/MULTICAST) IN EXISTENCE FOR THE LAST 5 YEARS .................... 44
4.12 POSSIBILITY OF UNDUE ADVANTAGE TO SOME PARTICIPANTS DUE TO
SEQUENTIAL ACCESS DUE TO EARLY LOG-IN ..................................................................... 46
4.13 REQUIREMENT OF MAIN SERVER and BACK UP SERVER? ................................................. 46

CHAPTER 5: ORDER TO TRADE RATIO…………………………………………………………….….47

5.1 INTRODUCTION .............................................................................................................................. 48

5.2 AVERAGE ORDER TO TRADE RATIO FOR ALL ACTIVE ALGO PARTICIPANTS IN THE
LAST 3 YEARS (YEAR-WISE BIFURCATION) .......................................................................... 48
5.3 ORDER-TO-TRADE RATIO FOR THE TOP 10 PARTICIPANTS (BY TURNOVER) ACROSS
THE LAST 3 YEARS (YEAR-WISE BIFURCATION) ................................................................. 48
5.4 ORDER-TO-TRADE RATIOS FOR EFFICIENT VS INEFFICIENT MEMBERS ACROSS THE
LAST 3 YEARS (YEAR-WISE BIFURCATION) .......................................................................... 49
5.5 TREND OF ORDER-TO-TRADE RATIO OVER THE LAST 3 YEARS ....................................... 50
5.6 EXCHANGE PRO-ACTIVENESS IN CONTROLLING THE ORDER-TO-TRADE RATIO ........ 51
5.7 PENALTIES OR STRUCTURES IN PLACE TO CURB HIGH ORDER-TO-TRADE RATIOS ... 52
5.8 PENALIZING REPEAT OFFENDERS OF HIGH ORDER-TO-TRADE RATIO ........................... 54
5.9 CAUSES OF A VERY HIGH ORDER-TO-TRADE RATIO ........................................................... 54
5.10 WHAT CATEGORY OF HFT ALGORITHMS COULD LEAD TO HIGH ORDER-TO-TRADE
RATIOS ............................................................................................................................................ 54

CHAPTER 6: REGULATORY FRAMEWORK…………………………………………………………...55

6.1 OVERVIEW OF REGULATORY CONCERNS ............................................................................... 56

6.2 MEASURES ADOPTED BY SECURITIES MARKETS REGULATORS IN DIFFERENT
COUNTRIES: ................................................................................................................................... 56
6.3 WHAT DOES IOSCO PRESCRIBE .................................................................................................. 61

CHAPTER 7: SEBI DISCUSSTION PAPER: STRENGTHENING OF THE REGULATORY

FRAMEWORK FOR ALGORITHMIC TRADING & CO-LOCATION………...……................65

xi
 7.1 MINIMUM RESTING TIME FOR ORDERS ................................................................................... 66
7.2 FREQUENT BATCH AUCTIONS (PERIODIC CALL AUCTIONS) ............................................. 70
7.3 RANDOM SPEED BUMPS OR DELAYS IN ORDER PROCESSING / MATCHING .................. 73
7.4 RANDOMIZATION OF ORDERS RECEIVED DURING A PERIOD (SAY 1-2 SECONDS) ...... 76
7.5 MAXIMUM ORDER MESSAGE-TO-TRADE RATIO REQUIREMENT ...................................... 78
7.6 SEPARATE QUEUES FOR CO-LOCATION ORDERS AND NON-COLO ORDERS (2
QUEUES)…………………………………………………………………………………………..81
7.7 REVIEW OF TICK-BY-TICK DATA FEED .................................................................................... 83

CHAPTER 8: SURVEILLANCE SYSTEM AT STOCK EXCHANGES………………………………….85

8.1 CURRENT SURVEILLANCE STRUCTURE AT NSE.................................................................... 86

8.2 SURVEILLANCE FUNCTIONS COVERED AT NSE .................................................................... 86
8.3 CURRENT SURVEILLANCE MECHANISM ON NSE AT INTRA-DAY BASIS FOR ALGO
TRADES BASED ON ORDER TO TRADE RATIO ...................................................................... 87
8.4 RISK MANAGEMENT MECHANISMS FOR ALGORITHMS EMPLOYED BY THE
EXCHANGE ..................................................................................................................................... 88
8.5 SEPARATE RISK MANAGEMENT MECHANISMS FOR HFT ................................................... 90
8.6 SURVEILLANCE MECHANISMS IN PLACE TO CATCH HARMFUL HFT ACTIVITIES ....... 90
8.7 DO SURVEILLANCE MECHANISMS CHANGE IN PERIODS OF HIGH VOLATILITY.......... 91
8.8 KILL SWITCHES .............................................................................................................................. 91
8.9 NEED BY EXCHANGES TO ENHANCE ITS SURVEILLANCE SOFTWARE/ PROGRAMS ... 92

CHAPTER 9: ALGO TRADING AND EXCHANGE APPROVALS……………………………………..93

9.1 PROCEDURE FOR GAINING ALGORITHMIC TRADING APPROVAL .................................... 94
9.2 PROCESS FOR FIRST TIME ALGO APPROVAL OR ADDITIONAL SOFTWARE FROM
DIFFERENT IT VENDOR: .............................................................................................................. 94
9.3 RULE FOR CHECKING ALGOS AT A LOGIC LEVEL BY THE EXCHANGE. ......................... 96
9.4 MECHANISM TO CAPTURE CHANGE IN ALGOS AT PARTICIPANTS END BY THE
EXCHANGE ..................................................................................................................................... 96
9.5 BEST PRACTICES INTERNATIONALLY FOR ANY MEMBER WANTING ACCESS TO
ALGORITHMIC TRAINING .......................................................................................................... 97
9.6 QUALIFICATION REQUIREMENT FOR MEMBERS INTENDING TO DO ALGO TRADING 98
9.7 EXCHANGE ENSURE ROBUSTNESS OF ALGORITHMS ON AN ONGOING BASIS……….98
9.8 ALGO TESTING LAB……………………………………………………………………………...99

xii
CHAPTER 10: CONCLUSION…………………………………………………………………………….100

REFERENCES……………………………………………………………………………………………..105

QUESTIONNARE…………………………………………………………………………………………107

xiii
 LIST OF TABLES

Table 1.1: Algo products and Strategies

Table 2.1: Client and proprietary contribution to turnover

Table 2.2: Composition of client and proprietary orders from Algo/ HF Trading

Table 2.3: Top 20 participants based on algorithmic trading turnover

Table 4.1: Segment wise percentage of Algo orders coming from co-location.

Table 4.2: VSAT, Leased line and MPLS connectivity cost

Table 4.3: Co-location Infrastructure cost

Table 5.1: Average of order-to-trade ratio of all the trading members for the last three financial years
Table 5.2: Order-to-Trade Ratio for the top 10 participants (by turnover) across the last three financial
years
Table 5.3: Average order-to-trade ratio for efficient and non-efficient members (NSE)

Table 5.4: Average order-to-trade ratio for efficient and non-efficient members (BSE)

Table 5.5: Year wise trend of order to trade ratio of ALGO trading members

Table 5.6: Penalty charges levied at member level for high Algo order-to-trade ratio(NSE)

Table 5.7: Fair usage charges categorised according to the execution efficiency (NSE)

Table 5.8: Penalty charges levied at member level for high Algo order-to-trade ratio (BSE)

Table 5.9: Fair usage charges categorised according to the execution efficiency (BSE)
Table 7.1: Average time between modifications for Nifty 50 and Nifty next stocks on the NSE Spot
market (2013)
Table 8.1: Monetary penalty levied in the last three financial years

Table 8.2: Order-to-trade ratio at market level (2016)

Table 8.3: Highest order-to-trade ratio (2016)

xiv
Chapter-1

ALGORITHMS

1
 Chapter 1
ALGORITHMS

1.1 INTRODUCTION TO ALGORITHMS

Algorithmic trading is a special type of trading which contains computerized logic from making
trading decision to transaction approval. Algorithmic trading became a popular tool after
technological progress transformed the way assets are traded.

Algorithmic trading is the use of programs and computers to generate and execute (large) orders
in markets with electronic access. Orders come from institutional investors, funds and trading
desks of big banks and brokers.

Algorithmic trading, or computer-directed trading, cuts down transaction costs, and allows
investment managers to take control of their own trading processes. The main objective of algo
trading is not necessarily to maximize profits but rather to control execution costs and market
risk.

Algorithms have become such a common feature in the trading landscape that it is unthinkable
for a broker not to offer them because that is what clients demand. These mathematical models
analyze every quote and trade in the stock market, identify liquidity opportunities, and turn the
information into intelligent trading decisions.

Why Algorithms
 Institutional clients need to trade large amounts of stocks. These amounts are often
larger than what the market can absorb without impacting the price.
 The demand for a large amount of liquidity will typically affect the cost of the trade
in a negative fashion (``slippage’’)
 Large orders need to be split into smaller orders which will be executed electronically
over the course of minutes, hours, day.
 The procedure for executing this order will affect the average cost per share,
according to which algorithm is used.
 In order to evaluate an algorithm, we should compare the average price obtained by
trading with a market benchmark (``global average’’ of the daily price, closing price,
opening price, etc)

1.2 TYPES OF ALGORITHMS

I. Agency Trading algorithms (execution algorithms)
Execution algorithm means to systematically split a larger order into many smaller orders based
on the available liquidity. The most commonly used algorithms are Time Weighted Average
Price (TWAP), Volume Weighted Average Price (VWAP) and Percent of Value (PoV).

The Time Weighted Average Price algorithmic strategy will break an order up into many
smaller equal parts and execute them during the trading day, normally at 5 minute intervals. The
Volume Weighted Average Price strategy will estimate the average volume traded for each 5-
minute interval the order is traded using historical trading information. The goal is to split the
order into smaller pieces based on an average weighted volume.
 Large Institutional clients want superior execution with minimal market impact for
large orders.

2
  Provide intelligence on executing large basket orders
 Some types of Agency Algorithms: Volume Weighted Average Price (VWAP), Time
Weighted Average Price (TWAP), aggressive, passive algorithms
 For example, Buy side firms execute large orders using various kinds of execution
algorithms

II. Proprietary Trading Algorithms

Proprietary trading (also "prop trading") occurs when a trader trades stocks, bonds, currencies,
commodities, their derivatives, or other financial instruments with the firm's own money, as
opposed to depositors' money, so as to make a profit for itself.
Proprietary Trading algorithms are typically used with the strategies that involve directional
bets on the markets – Net Long or Short depending on the market direction. Within this subset,
we have Momentum, Mean Reversion and Trend Following strategies. Besides, another popular
set of strategies called as Spread strategies or Market Neutral (both Long/Short) is also part of
this suit of algorithms.
Another subset of strategies involving proprietary trading uses a variety of Arbitrage strategies
such as index arbitrage, statistical arbitrage, merger arbitrage or volatility arbitrage.

a) Momentum based Strategies

 Assuming that there is a particular trend in the market. An algo trader follows
those trends. Further to the assumptions, the markets fall within the week. Using
stats to determine if this trend is going to continue or if it will change in the
coming weeks. Accordingly, algo trader will make the next move.
 Algo traders base their algorithmic trading strategy on the market trends which
is determined by using statistics.
 This method of following trends is called Momentum Based Strategy.
b) Arbitrage
 Assume that a pharma-corp is to be bought by another company, then the stock
price of corp could go up. This is triggered by the acquisition which is a corporate
event. If a trader is planning to invest based on the pricing inefficiencies that may
happen during a corporate event (before or after), then the trader is using an
event-driven strategy. Bankruptcy, acquisition, merger, spin-offs etc could be the
event that drives such kind of an investment strategy.
 These strategies can be market neutral and used by hedge fund and proprietary
traders widely.

3
 c) Statistical Arbitrage
 When an arbitrage opportunity arises because of misquoting in prices, it can be
very advantageous to algo trading strategy.
 Although such opportunities exist for a very short duration as the prices in the
market get adjusted quickly. And that’s why this is the best use of algorithmic
trading strategies, as an automated machine can track such changes instantly.
 Statistical Arbitrage algorithms take advantage of relative mispricing between 2
stocks.
d) Market Making
 A market maker or liquidity provider is a company, or an individual, that quotes
both a buy and a sell price in a financial instrument or commodity held in
inventory, hoping to make a profit on the bid-offer spread, or turn.
 Market making provides liquidity to securities which are not frequently traded
on the stock exchange. The market maker can enhance the demand-supply
equation of securities.
 Let’s assume there is a market maker, who buys for Rs. 500 from the market and
sell it at 505. He will give a bid-ask quote of Rs. 505-500. The profit of Rs. 5
cannot be sold or exchanged for cash without substantial loss in value.
 When the market maker takes a higher risk then the profit is also higher.

e) Machine Learning based

 In Machine Learning based trading, algorithms are used to predict the range for
very short term price movements at a certain confidence interval. The advantage
of using Artificial Intelligence (AI) is that humans develop the initial software
and the AI itself develops the model and improves it over time.
 A large number of funds rely on computer models built by data scientists and
quants but they’re usually static, i.e. they don’t change with the market. ML
based models on the other hand can analyze large amounts of data at high speed
and improve themselves through such analysis.

1.3 ADVANTAGES OF ALGORITHMS

a) Lower Costs
Algorithms are more cost effective for low-maintenance trades and that has meant head-count
shifts and reductions on sales desks. The ability to submit orders electronically to exchanges
directly rather than through brokers has been an important innovation in lowering the cost of
trading .Back-office functions and post-trade services such as clearing and settlement have also
benefited from automation.

b) Enables improved liquidity and pricing on shares

Algorithms are used extensively by broker-dealers to match buy and sell orders without
publishing quotes. By controlling information leakage and taking both the bid and offer sides
of a trade, broker Algorithms are in a way enabling improved liquidity, pricing on shares for
client, and higher commissions to brokers.

4
 c) Algorithm can analyse and react to the news faster before a human trader
An algorithm could, for example alert a trader if news is released on a company X and if the
company stock rises or falls by say one percent in the value of that stock within five minutes.
For example, Reuters News Scope Real-time product lets clients use live news content to drive
automated trading and respond to market-moving events as they occur. Each news item is ‘meta
tagged’ electronically to identify sectors, individual companies, stories or specific items of data
to assist automated trading.

d) To monitor and respond to risk conditions on real-time basis

Using real-time analytics, algorithms can continuously re-calculate metrics like Value-at-Risk
(VaR) and automatically hedge a position if VaR is exceeded.

e) Address Regulatory Compliance Issues

Compliance with law is of utmost importance and it is becoming burdensome with ever
increasing stringent regulations. Firms going forward will increasingly harness the latest in
algorithmic trading technology to address regulatory compliance issues.

f) Automate Surveillance
Regulators could automate surveillance to monitor algo-trading operations for patterns of
abuse.
However, limited availability of automated surveillance tools for algo trades and lack of skilled
staff and sufficient IT resources makes supervision technically challenging.

g) Offers Better Price with Minimal Market Impact

Most algorithms (categorized as execution algorithms) need to provide the best possible
execution price for clients hitting large orders in the market. Execution algorithms parcel a large
order into smaller orders into the market to minimize the market impact of a large order.
E.g. one of the most commonly used execution algorithms is implementation shortfall.
Implementation Shortfall is designed to measure and understand trading costs at the fund level,
by capturing the price slippage from the time the decision was made to implement an investment
idea to the time the trade was actually executed. Most algorithms already allow customers to
change the timing of executions, the rate of order-filling attempts at the beginning or end of the
trading day, and the tolerance for the slippage of a stock from certain benchmarks.
Other commonly used strategies are arrival price, time weighted average price (TWAP),
volume weighted average price (VWAP) and market-on-close (MOC).

1.4 DISADVANTAGES OF ALGORITHMS

a. Technical sufficiency and resources required

One of the biggest disadvantages of algo trading is the technical sufficiency and resources
required for algo trading. Algo trading requires knowing how to program in specific program
languages, which can take quite a while to learn. This facility may not be accessible to retail
investors and small traders.
b. Lack of Control
Since trades are automated, if the program runs in a way that one doesn't want it to, one will
be unable to control losses. Programs need to be tested thoroughly in order to avoid these
mistakes that might be made.

5
c. Lead to systemic risk
Interconnections between markets, which may be amplified by algorithms programmed to
operate on a cross-market basis, may allow for a shock to pass rapidly from one market to
another, potentially increasing the speed at which a systemic crisis could develop. This was
illustrated by the Flash Crash event of May 2010.

Facts of Flash Crash on 6 May 2010: That afternoon, major equity indices in both the futures
and securities markets, each already down over 4% from their prior-day close, suddenly
plummeted a further 5-6% in a matter of minutes before rebounding almost as quickly. Many
of the almost 8,000 individual equity securities and exchange traded funds (ETFs) traded that
day suffered similar price declines and reversals within a short period of time, falling 5%, 10%
or even 15% before recovering most, if not all, of their losses. Over 20,000 trades across more
than 300 securities were executed at prices more than 60% away from their values just moments
before. Moreover, many of these trades were executed at prices of a penny or less, or as high
as $100,000, before prices of those securities returned to their pre-crash levels. By the end of
the day, major futures and equities indices recovered to close at losses of about 3% from the
prior day.

d. Lack of Visibility
We know what a specific algorithm is supposed to do, measure its pre-trade analytics and see
how the post trade results match up to that expectation. But if the trader didn’t select the most
optimal algorithm for that trade little can be done. This problem is caused by a lack of visibility
and transparency into the algorithm while it is executing orders.

e. Algorithms Acting on Other Algorithms

If fund managers’ trading pattern is spotted and tracked with the use of algorithms, then these
algorithms are liable to be ‘reverse engineered’. This implies that their buy and sell orders are
pre-empted and used to the maximum effect by their competitors. Here, algorithms are acting
on other algorithms.

f. Lack of standard benchmark

With brokers offering many algorithmic strategies, one concern is that buy-side institutions lack
the tools to understand which algorithm to use for a particular stock. The lack of a standard
benchmark has made it almost impossible to assess the quality of algorithms.

g. Algorithmic trading requires careful real-time performance monitoring as well as pre

and post-trade analysis
This is required to ensure algo trade is properly applied. Traders must calibrate the algorithms
to suit portfolio strategy. More important is aligning execution choices with the level of order
difficulty involved in terms of: order size, liquidity, and trade urgency. Low touch venues such
as algorithmic trading lend themselves best to easier types of orders such as low-urgency and
small orders for large cap stocks. But urgent orders for a large volume of small cap stocks
would require a higher-touch approach to ensure best execution and cost efficiency.

High Touch Orders: Are orders that are sent electronically to the sell-side with the
expectation that a broker/trader will use discretion to add value to the execution process.

6
 Low Touch: Are electronic orders that are subjected to some form of a sell-side algorithm
(e.g. smart order routing) or the possibility of some form of human involvement that might
introduce latency into the execution process

h. Missing Ingredient—The Trader’s Gut Feel

Algorithms are simply advanced trading tools and they cannot replace the human elements or
make interaction redundant. Algorithms fail to capture a trader’s “gut feel”. It is the intraday
trading characteristics of a stock that assist a trader in determining the right strategy, whether
to back off or be more aggressive.

i. Fat finger/faulty algorithms - Any flaw in algorithms can cause huge deviations from the
healthy supply-demand arbitered prices
Algorithms react so quickly that by the time a human understands what is going wrong and
pulls the plug, hundreds of millions of rupees can be lost. Below is a list of some Flash Crash
Cases in India:

 Flash crash had occurred On BSE Muhurat Session in 2011. An algorithm that went
into a loop kept entering repeat trades in the Sensex futures contract, resulting in trades
worth ₹25,000 crore from one member in that session. All the trades in Sensex futures
of that session had to be annulled.

 On April 21st, 2012, the Nifty April futures plunged to 5,000 from 5,300 levels with
about 35,000 lots of Nifty futures getting traded in the space of a few minutes. The
sharp drop in futures also dragged the underlying index, with the 50-share nifty
declining from 5,313 to 5,245 within a few seconds. Nifty April futures finally closed
at 5,304.8, down 0.96 per cent; while the benchmark Nifty closed at 5,290.85, down
0.78 per cent. According to market buzz, the sell order as placed due to an algorithmic
trading error by a leading foreign institutional investor.

 In June 2010, the Reliance Industries stock had crashed nearly 20 per cent on execution
of a large ‘sell’ order using algo. The order, which appeared to be a punching error, saw
the Sensex plunge more than 600 points the moment it was executed.

1.5 CLASSIFICATION OF ALGORITHMS BASED ON LIFE-CYCLE STAGE

Algorithms are used extensively in various stages of the trading cycle. We can classify them
into pre-trade analytics, execution stage, and post-trade analytics.

Pre-Trade Analytics
The Pre-trade analytics involve thorough analysis of historical data and current price and
volume data to help clients determine where to send orders and when; whether to use algorithms
or trade an order manually we can call this as back testing the algorithm etc.,. The pre-trade
analysis is designed to help buy-side traders understand and minimize market impact by
choosing the level of aggressiveness and a time horizon for trading various stocks. Traders can
select varying levels of aggressiveness and visualize them against the time horizon for
completing the trade. Most compare the spread between bid and ask prices, reference that
against the volatility of a given stock, and attempt to create a range of potential outcomes. A
lot of the broker-sponsored algorithmic trading systems attempt to measure or project the trade
costs.

7
Execution
In the Execution stage, traders can create the lists of stocks, choose a particular strategy such
as implementation shortfall and enter the start time and the end time. Traders can also monitor
the performance and progress of the algorithms in real time and change the parameters if the
stock is moving away. Additionally, users can filter portfolios by sector, market cap, exchange,
basket, and percent of volume, profit and loss per share. Several brokers are designing
algorithms that sweep crossing networks and so-called dark books liquidity pools that match
buy and sell orders without publishing a quote.

Post-trade Analytics
Post-trade analytics track commissions and assist in uncovering the costs involved from the
time a trade is initiated all the way through to execution. Post-trade analytics are meant to
improve execution quality and facilitate the making of investment decisions. The most
prevalent trading benchmark in use today is VWAP, which is popular because it is easy to
measure. Although it provides comparative results, it is not as useful for evaluating strategies
that are trying to do something other than follow the market midpoint.

Following are the building blocks to creating algorithms:

8
1.6 LIFECYLE OF AN ALGORITHM
Following is a typical life cycle of an algorithm:

How do you start with the implementation of algorithmic trading strategies?

1. Decide upon the genre/strategy paradigm

The first step is to decide the strategy paradigm. It can be Market Making, Arbitrage based,
Alpha generating, Hedging or Execution based strategy. For this particular instance, let us
choose pair trading which is a statistical arbitrage strategy that is market neutral (Beta neutral)
and generates alpha, i.e. makes money irrespective of market movement.

2. Establish Statistical significance

You can decide on the actual securities you want to trade based on market view or through
visual correlation (in the case of pair trading strategy). Establish if the strategy is statistically
significant for the selected securities. For instance, in the case of pair trading, check for co-
integration of the selected pairs.

3. Build Trading model

Now, code the logic based on which you want to generate buy/sell signals in your strategy. For
pair trading check for “mean reversion”; calculate the z-score for the spread of the pair and
generate buy/sell signals when you expect it to revert to mean. Decide on the “Stop Loss” and
“Profit Taking” conditions.

o Stop Loss– A stop-loss order limits an investor’s loss on a position in a security. It fires
an order to square off the existing long or short position to avoid further losses and helps
to take emotion out of trading decisions.

9
 o Take Profit– take-profit orders are used to automatically close out existing positions in
order to lock in profits when there is a move in a favorable direction.

4. Quoting or Hitting strategy

It is very important to decide if the strategy will be “quoting” or “hitting”. Execution strategy to a
great extent decides how aggressive or passive your strategy is going to be.

o Quoting– In pair trading you quote for one security and depending on if that position gets
filled or not you send out the order for the other. In this case, the probability of getting a fill
is lesser but you save bid-ask on one side.
o Hitting- In this case, you send out simultaneous market orders for both securities. The
probability of getting a fill is higher but at the same time slippage is more and you pay bid-
ask on both sides.

5. Backtesting & Optimization

How do you decide if the strategy you chose was good or bad? How do you judge your
hypothesis?
This is where back-testing the strategy comes as an essential tool for estimation of the performance
of the designed hypothesis based on historical data. A strategy can be considered to be good if the
backtest results and performance statistics back the hypothesis.
Hence, it is important to choose historical data with a sufficient number of data points. This is to
create a sufficient number of sample trades (at least 100+ trades) covering various market scenarios
(bullish, bearish etc.). Ensure that you make provision for brokerage and slippage costs as well. This
will get you more realistic results but you might still have to make some approximations while
backtesting. For instance, while backtesting quoting strategies it is difficult to figure out when you
get a fill. So, the common practice is to assume that the positions get filled with the last traded price.

What kind of tools should you go for, while backtesting?

Since backtesting for algorithmic trading strategies involves a huge amount of data, especially if you
are going to use tick by tick data. So, you should go for tools which can handle such mammoth load
of data.

6. Risk and Performance Evaluation

No matter how confident you seem with your strategy or how successful it might turn out previously,
you must go down and evaluate each and everything in detail. There are several parameters that you
would need to monitor when analyzing a strategy’s performance and risk. Some important
metrics/ratios are mentioned below:

o Total Returns (CAGR) - Compound Annual Growth Rate (CAGR). It is the mean annual
growth rate of an investment over a specified period of time longer than one year.
o Hit Ratio- Order to trade ratio.
o Average Profit per Trade- Total profit divided by the total number of trades
o Average Loss per trade- Total loss divided by the total number of trades
o Maximum Drawdown– Maximum loss in any trade
o Volatility of Returns- Standard deviation of the “returns”
o Sharpe Ratio- Risk adjusted returns, i.e. excess returns (over risk free rate) per unit volatility
or total risk.

10
1.7 WHAT ARE EXECUTION ALGORITHMS AND HOW ARE THEY
DEVELOPED
a) Algorithmic trading refers to trade execution strategies that are typically used by fund managers
to buy or sell large amounts of assets.
b) They aim to minimise the cost of these transactions under certain risk and timing
constraints. Such systems follow preset rules in determining how to execute each order. These
rules are pre-defined and coded in the form of algorithms and fed into a computer system.
c) Algorithmic trading systems are offered by many brokers and simply execute the orders that
they are given. Their job is to get a good price (as compared to various benchmarks) and
minimise the impact of trading. This is done by slicing orders and dynamically reacting to
market events.

11
Agency algorithms at I-Banks:

 All global I-bank have a desk in algorithmic trading

 Risk savvy banks -higher Prop focus E.g.: GS (US Banks)
 Risk averse banks -higher Agency focus E.g.: CS (EU Banks)

Table 1.1: Algo Products and Strategies

12
1.8 WHAT ARE HFT ALGORITHMS AND HOW ARE THEY DEVELOPED
d) High-frequency trading (HFT) is a subset of automated trading. Here, opportunities are sought
and taken advantage of on very small timescales from nanoseconds up to milliseconds.
e) Some high-frequency strategies adopt a market maker type role, attempting to keep a relatively
neutral position and proving liquidity (most of the time) while taking advantage of any price
discrepancies.
f) Other strategies invoke methods from time series analysis, machine learning
and artificial intelligence to predict movements and isolate trends among the masses of data.
g) Specifics of the strategy aside, for HFT, monitoring the overall inventory risk and incorporating
this information into pricing/trading decisions is always vital.

13
 Chapter-2

COMPOSITION OF ALGORITHMIC
TRADING

14
 Chapter 2
COMPOSITION OF
ALGORITHMIC TRADING

2.1 COMPOSITION OF ALGO TRADING TAKING PLACE IN THE COUNTRY-

CLIENT OR PROPRIETARY

Client and proprietary contribution to turnover for all algorithmic orders across segments (CM, F&O
and CDS) for the period FY 16-17 (Apr’16 to Feb’17) is provided below:

Table 2.1: Client and Proprietary Contribution to Algo Turnover

Category % to Exchange Algo Turnover

Client 58%
Proprietary 42%
Notes:
1. In case of options, premium turnover is considered in F&O and CD segment.
2. Algorithmic order is identified based on identification code as provided by the trading member.
3. Proprietary account is identified where the member code is equal to client code.

Both Client and Prop orders are received from Algo/HFT. The composition of the same for FY 2016
till now is a given in the below table.

Table 2.2: Composition of Client Proprietary Orders from Algo/HFT

SEGMENT PROPERIOTRY % CLIENT %

EQUITY 34.00 66.00
EQUITY DERIVATIVES 76.05 23.95
CURRENCY DERIVATIVES 63.38 36.62
INTEREST RATE DERIVATIVES 32.97 67.03

2.2 CURRENTLY REGISTERED ALGORITHM TRADING PLAYERS

A) Currently, there are 251 trading members as on February 2017 registered for Algo trading with
the Exchange.
B) 233 trading members are active.
Note: Trading members with at least one trade during the period from Dec 2016 to Feb 2017 in Capital
Market (CM) or Futures and Options (F&O) or Currency Derivatives (CD) segment have been
considered as active.

15
As on Feb 28, 2017, 141 trading members have taken approvals for Algo Trading facility out of which
35% i.e. 52 trading members are active.

2.3 TOP 20 ALGO PARTICIPANTS AND THEIR DAILY TURNOVER

Top 20 participants (trading members) ranked based on algorithmic trading turnover irrespective of
colocation or non-colocation across segments (CM, F&O and CDS) for the period FY 16-17 (Apr’16
to Feb’17) and their contribution to total Exchange turnover is provided below:

Table 2.3: Top 20 participants ranked based on Algorithmic Trading Turnover

Participants Rank % Contribution to

(Basis Algo Turnover) Exchange Turnover
1 6.02%
2 4.37%
3 2.14%
4 1.80%
5 1.66%
6 1.43%
7 1.37%
8 1.19%
9 1.17%
10 1.11%
11 1.07%
12 1.06%
13 0.97%
14 0.95%
15 0.92%
16 0.85%
17 0.78%
18 0.69%
19 0.67%
20 0.64%
TOTAL 30.86%
Notes:
1. In case of options, premium turnover is considered in F&O and CD segment.
2. Algorithmic order is identified based on identification code as provided by the trading member.

16
2.4 ALGORITHMIC TRADING TRENDS AND THE EXTENT TAKING PLACE
IN INDIAN MARKETS VIS-À-VIS GLOBAL MARKETS
• After the initial spurt the share of Algorithmic trading to Exchange turnover has stabilised
around 47% in India across cash and derivatives segment on NSE.

Source: NSE

 As can be observed from below, majority of the trading activity of algo players is only in liquid
scrips/contracts

• In developed markets such as US, it stands at approximately 70 - 80%.

17
US Equity Markets

Source: Morton Glantz, Robert Kissell. Multi-Asset Risk Modeling: Techniques for a Global Economy
in an Electronic and Algorithmic Trading Era.

Asian Markets

18
2.5 COMPARISON OF INDIAN MARKETS VIS-À-VIS GLOBAL MARKETS

Global Markets Indian Markets

Exchanges, Alternate Trading Systems, Dark Pools Registered Exchanges

Price Time Priority Retail Clients about 93%

Internalizing, Preferencing, Routing
orders limited orders

Orders to pass through member risk management All orders pass through the member order
system level risk management systems

Complex web of interconnectivity amongst

Only two Stock Exchanges
multiple venues (40 plus)

Market Data free for all members and its

Market Data only through vendors
investors

All orders pass through member order level

Payment for order flow
risk management systems

Maker Taker Pricing No such facility

Servers hosted in India. Audit trail for all

Broker Conflicts
orders generated by Algo

19
 Chapter-3

HIGH FREQUENCY TRADING (HFT)

20
 Chapter 3
HIGH FREQUENCY TRADING (HFT)

3.1 INTRODUCTION TO HFT ALGORITHMS

a) High-frequency trading (HFT) is a subset of automated trading. Here, opportunities are sought
and taken advantage of on very small timescales from nanoseconds up to milliseconds.
b) Some high-frequency strategies adopt a market maker type role, attempting to keep a relatively
neutral position and proving liquidity (most of the time) while taking advantage of any price
discrepancies.
c) Other strategies invoke methods from time series analysis, machine learning
and artificial intelligence to predict movements and isolate trends among the masses of data.
d) Specifics of the strategy aside, for HFT, monitoring the overall inventory risk and incorporating
this information into pricing/trading decisions is always vital.

3.2 BENEFITS OF HFT

a) Lower Costs (lower bid-ask spreads)

Technological innovation, together with financial deregulation, has led to a long-term decline
in trading costs in many financial markets and these benefits have been shared by both financial
intermediaries and consumers of financial services.
The benefits of automation are not confined to trade execution. Back-office functions and post-
trade services such as clearing and settlement have also benefited from automation.

The ability to submit orders electronically to exchanges directly rather than through brokers has
been an important innovation in lowering the cost of trading.

Market Evidence

i. In the US, equity market quality and liquidity have improved alongside the rise of
electronic trading on a wide-range of measures. With HFT accounting for as much as
70% of US equity market turnover, the US also enjoys the world’s lowest institutional
trading costs for large cap stocks.
ii. In Australian equity markets, institutional brokerage and transaction costs have been
on a declining trend in recent years, consistent with overseas trends. Below diagram

21
 shows average commission and implementation shortfall (the difference between
arrival and execution prices) in Australian equity markets measured in basis points as
compiled by ITG, an independent broker-dealer.
iii. A similar level and trend for implied institutional commission rates is evident in a
survey of Australian fund managers by Peter Lee Associates (see below diagram).
While the commissions and other transaction costs paid by Australian fund managers
are influenced by a range of factors apart from the growth in HFT, the declining trend
in these costs is consistent with the view that HFT lowers costs for investors.

Institutional Transaction Costs-Australia Basis Points

Average Commission/Average Portfolio Size Basis Points

b) Increased Liquidity
HFT provides liquidity in markets, which is essential to their functioning. Liquidity can be
defined as the ease with which market participants can buy or sell financial instruments or
securities. It is usually measured by reference to the spread between buying and selling prices,
although measures of market depth (the volume of a security on bid or offer), the time taken to
complete a trade (immediacy), total turnover and turnover ratios are also relevant measures of
market liquidity. By adding to market liquidity, HFT narrows buy-sell spreads and lowers the
cost of trading. HFT has also been important in driving reductions in tick sizes, which facilitates
the narrowing in spreads.
HFT imposes a positive externality on market participants, because HFT firms do not fully
capture all of the benefits of HFT in lowering trading costs for other market participants, nor
do they fully capture the social value of the information generated through price discovery in
financial markets.

22
 c) Increased Efficiency
By lowering transaction costs, HFT improves the efficiency of markets. The main function of
financial markets is price discovery, which in turn coordinates the economy-wide capital
allocation process. The efficiency of financial markets can be defined as the speed with which
markets incorporate new information in asset prices. The increased velocity of trading through
HFT ensures that market prices reflect new information more quickly. Much of the innovation
in financial markets historically has been driven by the desire to profit from bringing
information to market more quickly.

d) Higher Returns for Investors

Transaction costs subtract from returns to investors. Lowering transaction costs raises returns
to investors and thus asset prices. It follows that any permanent reduction in transaction costs
through innovations such as HFT will lead to a permanent increase in asset prices and a positive
wealth effect from the increased value of investors’ portfolios.

e) Lower Cost of Capital

Higher asset prices also lower the cost of capital for firms, increasing investment, the capital
stock, productivity, real wages and living standards. It is said, ‘this is the main channel by which
HFT can have societal value.

f) Profitability and Market Efficiency

As financial markets become more efficient, the profitable trading opportunities available to
HFT firms are reduced. As in any other industry, the competitive entry of HFT firms exploiting
new technologies can be expected to reduce profits available to HFT and algorithmic trading
over time. At some point, the marginal cost of deploying new HFT technologies will exceed
the marginal benefit available from HFT-based trading strategies. Trends in the profitability of
HFT firms may be indicative of the efficiency of financial markets.

g) Lower Volatility
Empirical evidence overwhelmingly supports the conclusion that HFT enhances market quality
and reduces market volatility. HFT smooths market prices by trading against transitory price
changes and in the direction of permanent price changes. There is also evidence that HFT
reduces the probability of end-of-day price or market manipulation. These are all natural
consequences of the role of HFT in improving market liquidity and efficiency.

3.3 DEMERITS OF HFT

HFT Impact on Retail Investors and Small Traders

a) HFT is playing against investors who are willing to invest fundamentally

Many HFT algorithms are based on finding a broker which is starting to purchase shares, than
utilizing the fact that HFT Company has quicker links to other brokers - promptly buy available
shares from that broker, and re-sell them to the one willing to buy at higher price. This is making
the market less attractive to fundamental investors. It is a kind of a hidden tax for fundamental
players. However this 'con' is smaller than it used to be because the majority of HFT
transactions are done against other HFT's.

b) HFT firms leverage special services such as co-location facilities and raw data feeds

23
 These facilities are typically not accessible to smaller firms and retail investors as they are not
able to make the required investments. This places these smaller firms and investors at a
disadvantage as they cannot afford these facilities.

c) Some HFT firms often enter trades just for the liquidity rebate
In rebate trading, instead of trader paying the commission for buying and selling, he is being
paid by the service provider. ECN (Electronic Communication Network) rebate is the primary
source of profit. This adds no value to the retail or long-term investor

d) Short term investors’ end up paying higher price

Both traders and short-term investors who invariably whirl in and out of financial securities
are victimized in that they end up paying a higher price over a short window or limited time for
what they buy. However for long term investors, the impact is much less and not as dramatic
on their total returns since they don’t depend on quick returns.

e) HFT firms gather certain market information faster

Some analysts believe the fact that HFT firms are often able to gather certain market
information and execute trades faster than other participants puts the overall fairness and
integrity of the markets at risk.

3.4 HFT IMPACT ON INSTITUTIONAL INVESTORS

a. High-frequency traders detect and front-run large orders by institutional investors

This is a practice that pushes the stock price up (down) if institutional investors have large buy
(sell) orders (i.e. upon detecting an incoming order flow, after which the HFT system buys the
same security and then turns around and sells it to the institution at a slightly higher price)
thereby also increasing stock price volatility.

b. HFT leverage services such as Colocation

HFT firms co-locate their computers physically close to the exchanges’ computers to gain
millisecond speed advantages, so they can beat slower orders from buy-side institutional
investors to the quote. In some cases, HFT firms even co-locate their computers in the same
room as an exchange’s computers

c. Impact on Shortlfall
Shortfall is the difference between the prevailing price or value when a buy or sell decision is
made with regard to a security and the final execution price or value after taking into
consideration all commissions, fees and taxes. As such, implementation shortfall is the sum of
execution costs and the opportunity cost incurred in case of adverse market movement between
the time of the trading decision and order execution.

Market Evidence
A paper published by Professor Lin Tong “A blessing or a curse? The impact of High
Frequency Trading on Institutional Investors” (2015) (attached below) suggests that high
frequency trading often has a negative effect on institutional trading. His findings are as
follows:

24
  HFT activity is positively correlated with execution shortfall
 When HFT activity is more intense, institutional investors’ execution shortfall is
higher.
 The increasing effect of HFT activity on execution shortfall is stronger on smaller
stocks.
 When HF traders on the net are buying (selling), it is more costly for institutional
investors to sell (buy).
 Even though HFT activity increases institutional investors’ execution shortfall, it does
not provide the benefit of reduced timing delay cost.

Research Paper

lin_tong HFT on
institutional investors.pdf

d. Mutual funds are worst hit

Mutual funds (mostly active funds) that do a lot of large block trading and tend to trade
urgently or predictably are worst hit. Since mutual funds trade in large quantities, there are
possibilities where other HFT algorithms sniff these large orders and capitalize on this
information, thus front-running large orders and manipulating markets in the process.

e. HFT strategies use the order flow entering the market to extract profits, whereas
institutional investors use fundamental information to extract profits

HFTs profit from the order flow information arriving from the institutional investors.
Essentially, HFTs are skimming cream off the top of an institutional investor’s fundamental
analysis, but they win regardless of whether the institutional investor’s analysis was correct
or not.

f. HFT hinders the market’s ability to incorporate news about a firm’s fundamentals
into asset prices

Market Evidence

Froot, Scharfstein, and Stein (1992) Research Paper: “Herd on the Street: Informational
Inefficiencies in a Market with Short-Term Speculation” (attached below)

Using analyst forecast revisions and earnings surprises as proxies for news about firm
fundamentals; it is found that stock prices react more strongly to news about fundamentals
when HFT is at a high volume. However, the incremental price reactions due to HFT are
almost entirely reversed in the subsequent period. Taken together, the evidence suggests
that HFT exaggerates otherwise-sound price reaction. The price swings introduced by HFT
also represent direct evidence that HFT increases stock price volatility.

25
 Research Paper:

herd_on_the_street
_froot scharfstein.pdf

g. HFT firms are often able to gather certain market information and execute trades
faster

This puts other participants at a disadvantage and the overall fairness and integrity of the
markets at risk.

h. Increasing cost of Transaction

The temporary price impact of large trades causes noise in prices due to price pressure
arising from liquidity demand by long-term investors. If HFTs trade against this transitory
pricing error, they can be viewed as reducing long-term investors’ trading costs.

However, if HFTs trade in the direction of the pricing error, they can be viewed as
increasing the costs to those investors. HFTs trading in the direction of pricing errors could
arise from risk management, predatory trading, or attempts to manipulate prices while
HFTs following various arbitrage strategies could lead to HFTs trading in the opposite
direction of pricing error

3.5 HFT IMPACT ON VOLATILITY

a. HFT is positively correlated with stock price volatility after controlling for the volatility
of a stock’s fundamentals and other volatility drivers.
There seems to be a positive correlation between HFT and volatility that grows stronger during
periods of high market uncertainty. Hence, a certain section of the market believes that HFT
increases volatility.

Evidence: Research Paper by Zhang (2010) on The Effect of High-Frequency Trading on Stock
Volatility and Price Discovery (research paper attached below)

Study shows that this positive correlation is especially strong for the top 3,000 stocks in market
capitalization and for stocks with high institutional holdings. The positive correlation between
HFT and volatility is also stronger during periods of high market uncertainty. Taken together,
the results are consistent with the view that HFT increases volatility

Research Paper:

Zhang effect of hft

on volatility.pdf

26
b. HFT hinders price discovery
HFT causes stock prices to overreact to news about company fundamentals. The incremental
price changes due to HFT are almost entirely reversed in the subsequent periods. Thus, HFT
algorithms could cause erratic movements in markets, thus hindering price discovery.

c. HFT can give rise to price fluctuations and short term volatility
HFT involves rapid intraday trading with positions generally held only for minutes—or even
just seconds. Given that HFT volumes are normally a relatively high percentage of overall
trading; the price fluctuations caused by this strategy can lead to overall volatility in the market.

Also, the practice of making trades and instantly cancelling them only to trigger automated
buying from other firms is an ethical issue that has been questioned by many analysts.

 HFT algorithms have made certain market related jobs redundant

High frequency trading has displaced many of the traditional market makers whose job it was
to keep an orderly market, and who were required to step in and provide a backstop during
volatile periods.

 The odds of a sudden liquidity drain go up

HFT firms are under no obligation and can enter or leave the market unbelievably fast, the odds
of a sudden liquidity drain go way up.

Market Evidence:
This is what we witnessed during the flash crash in US markets on May 6, 2010. One mutual
fund firm entered an ETF order incorrectly, and in a matter of minutes that information had
flashed across the market and without human intervention, liquidity was instantly pulled. Some
stocks were left with no bids at all and traded at $0.01. The market recovered later in the day,
but the damage was done.

Partly in response to these calamities, regulators have introduced circuit breakers into the
market that automatically halt trading if a stock moves more than a certain amount. The hope
is that these algorithmic halts will give human traders time to intervene and prevent widespread
disruption.

 Fat finger/faulty algorithms - Any flaw in algorithms can cause huge deviations from the
healthy supply-demand arbitered prices
In a real market instance of machines running amok, on August 1st, 2012 Knight Capital
mistakenly put a trading algorithm meant for development into the US markets and in a matter
of an hour or two the system had racked up so many losing trades that the company almost went
bankrupt. These programs react so quickly that by the time a human understands what is going
wrong and pulls the plug, hundreds of millions of dollars can be lost.

Flash Crash Cases in India

Flash crash had occurred On BSE Muhurat Session in 2011. An algorithm that went into a loop
kept entering repeat trades in the Sensex futures contract, resulting in trades worth ₹25,000

27
 crore from one member in that session. All the trades in Sensex futures of that session had to
be annulled.
On April 21st, 2012, the Nifty April futures plunged to 5,000 from 5,300 levels with about
35,000 lots of Nifty futures getting traded in the space of a few minutes. The sharp drop in
futures also dragged the underlying index, with the 50-share nifty declining from 5,313 to 5,245
within a few seconds. Nifty April futures finally closed at 5,304.8, down 0.96 per cent; while
the benchmark Nifty closed at 5,290.85, down 0.78 per cent. According to market buzz, the sell
order as placed due to an algorithmic trading error by a leading foreign institutional investor.

In June 2010, the Reliance Industries stock had crashed nearly 20 per cent on execution of a
large ‘sell’ order using algo. The order, which appeared to be a punching error, saw the Sensex
plunge more than 600 points the moment it was executed.

3.6 MISUSING HFT FOR MARKET MANIPULATION

Harmful HFT Algorithms that might cause market disruption such as high Intra-Day volatility
and high Order-to-Trade Ratio.

A lot of academic research has happened on algorithmic trading. Some of the inferences from the
research paper are given below:-

a) Susan Thomas and Nidhi Agarwal from IGIDR have attempted to study the impact of
algorithmic trading on market quality and has published their research paper titled “The causal
impact of algorithmic trading on market”. Their findings suggest that algorithmic trading
improves market quality. There are improvements in transactions costs, volatility, and buy-sell
imbalance. There are improvements in some, but not all of the depth measures and these are
sensitive to the match design. Two areas where the results provide new insights are the intraday
volatility of liquidity and the probability of extreme price movements and reversal over a very
small period during the day, often referred to as a flash crash. Policy makers have been very
concerned that liquidity provided by algorithmic trading can rapidly deteriorate when news
breaks. Their results show that the liquidity risk is lower with more algorithmic trading. A
similar concern has often been voiced about the probability of a flash crash. However, they
found that higher algorithmic trading intensity either leads to fewer of such episodes or has no
effect. They have concluded by inferring that there are more benefits than costs to securities
that attract higher algorithmic trading activity. With proper safeguards in place, more
meaningful policy measures could be built to increase the level of algorithmic trading to a
broader base of securities, rather than inhibit it.

b) In another research paper titled “How do High-Frequency Traders Trade?” By Nupur

Pavan Bang and Ramabhadran S. Thirumalai (ISB) have tried to examine the order handling
and trading behavior of high-frequency traders (HFTs) around firm specific earnings surprises
as well as unexpected interest rate changes. In this paper they have observed that high frequency
traders do not significantly change their order handling and trading behavior around earnings
surprises but do reduce their participation in the market after a macroeconomic shocks. High
frequency traders also do not change the mix of order types that they use around earnings
surprises, although they appear to increase the number of liquidity-demanding orders. They use
fewer aggressively priced limit orders around macroeconomic shocks. The profitability of their
orders does not change around earnings announcements. However, despite reducing their
market participation after a macroeconomic shock, they still make losses to informed traders.

28
 The research work also tried to examine the trading behavior of buy-side algo. They observed
an increase in their market participation around earnings surprises and do not change it around
macroeconomic surprises.

c) Austin Gerig From Division of Economic and Risk Analysis U.S. Securities and Exchange

Commission has published a paper titled “High-Frequency Trading Synchronizes Prices in

Financial Markets”. The paper has inferred that high frequency trading facilitates information
transfer between investors, which increases the accuracy of prices and redistributes profits from
informed individuals to average investors by reducing transaction costs.

d) European Central Bank has published a working paper titled “High Frequency Trading and
Price Discovery”. The paper in its concluding remarks have stated that high frequency trading
increase the efficiency of prices by trading in the direction of permanent price changes and in
the opposite direction of transitory pricing errors. Their results also show no direct evidence
that high frequency traders contribute directly to market instability in prices.

3.7 MARKET MANIPULATION TECHNIQUES USING HFT

I. QUOTE STUFFING
Quote stuffing is a practice where a large number of orders to buy or sell securities are placed
and then cancelled almost immediately. It is estimated that the majority of Quote stuffing
episodes last up to 2 seconds (as per research by Tse, Lin, and Vincent (2012))

Quote stuffing is easily observed via several characteristic patterns of quote volatility which
may occur on the Ask, the Bid or both simultaneously. Empirically, Quote Stuffing is observed
as having some influence on the direction of price moves immediately following an episode
with prices seen as more likely to move in the direction of the stuffing.

Evidence 1

Research Paper by Egginton and Van Ness (2011) “Quote Stuffing” examines the impact of
intense episodic spikes in quoting activity (frequently referred to as quote stuffing) on market
conditions. It is found that that quote stuffing is pervasive and that over 74% of US exchange
listed securities experience at least one episode during 2010.

It is also found that stocks experience decreased liquidity, higher trading costs, and
increased short term volatility during periods of intense quoting activity. Also the most quote
stuffing events occur on the NYSE, ARCA, NASDAQ and BATS.

Research Paper:

quote-stuffing.pdf

29
Evidence 2

(Using Research Paper-High Frequency Trading – Measurement, Detection and Response by

Tse , Lin and Vincent(2012))

Tse, Lin, and Vincent (2012) note that during quote stuffing events, the trade prices tend to
move in the direction of the stuffing activity i.e. increase when stuffing occurs on the ask
side or decline when it is happening on the bid side.

On average, the price tends to move toward quote stuffing after the event (i.e. the mid-price
moves up if quote stuffing occurred on the offer). This holds whether the affected quote finished
“ticked in” – narrower than the initial spread – or “ticked out”, but is more pronounced when
finishing “ticked in” (see below diagram). However, these moves tend to be very small (<
0.23bps).

Average Mid-price move toward quote stuffing (5 seconds post event)

A stock which has already experienced quote stuffing has a higher probability of further activity
on that same day, with an 82.3% chance of a “repeat” event. The second event occurs on the
same venue 73% of the time, and over 70% of “repeats” occur within 5 minutes (see below
diagram)

Time between Quote Stuffing Events

30
II. LAYERING
i. Layering in Stock Trading:
This is a scheme used by securities traders to manipulate the price of a stock ahead of
transactions that they wish to execute, creating more advantageous executions for them. It is a
variety of a stratagem that has come to be called spoofing, itself an element of high frequency
trading.

Through layering, a trader tries to fool other traders and investors into thinking that significant
buying or selling pressure is mounting on a given security, with the intent of causing its price
to rise or fall.

The trader does this by entering multiple orders that he has no intention of executing, but instead
plans to cancel.

Buying Example: A trader is looking to buy 1,000 shares of XYZ stock, which is trading at
$20.00 per share. In hopes of pushing its price down, he enters 4 large orders to sell:

10,000 shares $20.05 per share

10,000 shares $20.10 per share
10,000 shares $20.15 per share
10,000 shares $20.20 per share

The trader has layered these sell orders at incrementally higher prices above the current market
price. Thus, they will not execute unless the current market price moves upward. The trader
intends to make other market participants believe that selling pressure is mounting among
holders of XYZ stock, and that the price thus is bound to fall below $20.00 per share.

Intention: If the scheme works, other traders eager to sell will enter orders below $20.00,
anticipating that those orders to sell 40,000 shares soon will be re-entered at even lower prices.

The trader then will be able to purchase 1,000 shares of XYZ at less than $20.00 per share, and
cancel those layered sell orders.

Risks Involved: The trader runs a risk that orders to buy XYZ will intervene, instead pushing
the price above $20.00 per share. In this case, the trader will have to deliver up to 40,000 shares
to buyers, shares that he might have to obtain at yet higher prices, incurring a large loss in the
process.

ii. Regulatory Response

The Dodd-Frank Financial Reform Bill of 2010 made all forms of spoofing illegal in the United
States. Under its provisions, for example, the U.S. Justice Department charged a U.K.-based
day trader with illegal actions that allegedly caused the May 6, 2010 “Flash Crash” in which
stock market prices plummeted suddenly.

Meanwhile, the SEC has taken enforcement actions against traders and firms who engaged in
spoofing and layering even prior to the passage of Dodd-Frank.

31
 iii. Market Evidence
(Using Research Paper-High Frequency Trading – Measurement, Detection and Response by
Tse, Lin and Vincent (2012))
Below exhibit shows a real example where a trade of [email protected] in Legrand SA on Euronext
Paris lead to the 1200 shares at 29.135 being cancelled within milliseconds. This behaviour can
impact performance and fill rates, particularly for aggressive trading that targets multiple levels
of displayed liquidity.

Price Fade Example, Legrand SA (Paris) September 21st, 2012

III. MOMENTUM IGNITION

Introduction
Momentum Ignition is an HFT strategy that is characterized by a specific pattern observed in
both trade prices and trade volumes. This usually comprises three main stages:
• An initial spike in trade volume, which is not accompanied by any significant changes
in price.
• A subsequent sharp price move (positive or negative), accompanied by a new, even
larger increase in volume
• A gradual price reversal to levels observed before the event, accompanied by low
volume.

This pattern is observed over significantly longer time frames than Quote Stuffing, and may
last for up to several minutes.

Evidence
Evidence 1 (using Research Paper-High Frequency Trading – Measurement, Detection and
Response by Tse , Lin and Vincent(2012))

Momentum ignition does not occur in the blink of an eye, but its perpetrators benefit from an
ultra-fast reaction time. Generally, the instigator takes a pre-position; instigates other market
participants to trade aggressively in response, causing a price move; then trades out. Momentum
ignition is identified with a combination of factors, targeting volume spikes and outsized price
moves - see below exhibit for an example of this pattern in Daimler on 13th July, 2012:

Daimler AG, 13th July, 2012

32
Though one cannot conclusively determine the intention behind every trade, this is the kind of
pattern one would expect to emerge from momentum ignition. This can be used as a proxy to
estimate the likelihood and frequency of these events.

Likelihood and Rapid Price Moves

As shown in below figure, it is estimated that momentum ignition occurred on average 1.6 times
per stock per day for STOXX 600 names in Q3 2012, with almost every stock in the STOXX600
exhibiting this pattern on average once a day or more.

Frequency of Momentum Ignition Patterns

In addition, it is noted that the average price move is 38bps (but over 5% are more than 75bps,
with some significantly higher – see below Exhibit), and the time it takes for that move to occur
is approximately 1.5 minutes (see below exhibit).

Price Move Distribution (Post Momentum Ignition Pattern)

Duration Distribution (Momentum Ignition Patterns)

33
 While 38bps may not sound like a big move, it is a bit more significant when compared to the
average duration of these events (1.5 minutes) and the average spread on the STOXX600
(approximately 8bps).

Though not all momentum ignition events result in massive price moves, those that do can
cause significant impact. Percentage of volume orders that would normally execute over hours
may complete in minutes on the back of “false” volume (one of the causes of the 2010 flash
crash was a straightforward percentage of volume order).

Sources:

1. An Empirical detection of High Frequency Trading strategies by Dimitar Bogoev1 and Arz´e
Karam of Durham University
2. High Frequency Trading – Measurement, Detection and Response by Tse , Lin and
Vincent(2012)

Research Papers:

HFT_Measurement_ hft strategies nd

Detection_and_Response_Tse
impact.pdf
Lin Jonathan.pdf

IV. LIQUIDITY REBATE TRADING

Introduction
Certain exchanges and ECNs offer traders a rebate for providing the markets with shares, or
liquidity, when there i
s a need for it. They generally pay traders to buy at the bid price and sell at the ask price, thereby
providing liquidity, and charge traders for placing market orders.

Liquidity rebate traders look for large orders, fill a part of that order, and then offer these shares
back to the market by placing a limit order, which makes them eligible to collect the rebate fee
for providing liquidity, with or without them making a capital gain.

Research Papers:

CapgeminiHigh_Freq
uency_Trading__Evolution_and_the_Future.pdf

34
Chapter-4

CO-LOCATION

35
 Chapter 4
CO-LOCATION

4.1 WHAT IS CO-LOCATION

In a sign of the rush for speed in trading, exchanges are building huge data centres where traders,
members and non-members alike, can place computers containing their trading algorithms next to an
exchange’s matching engine, which matches “buy” and “sell” orders. This “co-location” shaves crucial
milliseconds from the time it takes to complete a trade.

Basically, colocation is locating computers owned by HFT firms and proprietary traders in the same
premises where an exchange’s computer servers are housed. This enables HFT firms to access stock
prices a split second before the rest of the investing public.

If traders are located 100 miles away from an exchange, they face a delay of one millisecond whenever
they seek to trade a price via their computer screen. Few serious investors can afford to be that late to
prices that flash so quickly. Many HFT traders now operate in the smaller realm of microseconds.

Co-location has become a lucrative business for exchanges, which charge HFT firms by rack space for
the privilege of “low latency access.”

4.2 ADVANTAGES OF CO-LOCATION

a. Co-location reduces latency

Latency is the time it takes to make and execute trading decisions and therefore enables co-
located market-maker high-frequency traders to more rapidly adjust their quotes as market
conditions change.

b. Levels the playing field among competing HFT market makers

It also levels the playing field among competing HFT market makers who are also co-located
by ensuring that none has a latency advantage over the other from an exchange perspective.

c. Increases Liquidity
The introduction of co-location facilities is expected to increase liquidity by encouraging HFT
and increasing competition among HFT market makers.

Market Evidence
The ASX introduced co-location for ASX futures markets on February 20, 2012. As per the
research paper by Frino Mollica Webb (2013) “The Impact of Co-Location of Securities
Exchanges’ and Traders’ Computer Servers on Market Liquidity”, (attached below) there is
evidence of an increase in HFT activity following the introduction of colocation.

There is also strong evidence of a decrease in bid-ask spreads and an increase in market depth
following the introduction of colocation, across all futures contracts examined. It is also proved
that the introduction of colocation resulted in an improvement in liquidity of futures contracts.

36
Research Paper:

Co-Location_impact_
_Securities_Exchanges’_and_Traders’Computer_Servers_on_Market_Liquidity.pdf

4.3 DISADVANTAGES OF CO-LOCATION

a. Expensive
Colocation services can prove to be extremely expensive, especially if the HFT algorithms
work across multiple exchanges or locations

b. Market Inequity
HFT firms leverage special services such as co-location facilities, which are typically not
accessible for smaller firms and retail investors as they are not able to make the required
investments. This causes inequity and a huge disadvantage to smaller traders/retail investors.

4.4 CO-LOCATION FACILITIES IN INDIA

India boasts of the fastest colocation in the world. BSE is the fastest Exchange in the world with a
speed of 6 microseconds. Market access across Equity, Equity Derivatives and Currency
Derivatives segments.

BSE provides the fastest Co-location service in India with round trip network latency of less than
10 microseconds. At BSE Colocation response for an order has round trip latency of about 16
microseconds (including 10 microseconds of Co-location network latency).

AT BSE, there is equitable distribution of market data to all members by usage of same length
cables for all members. Also, Full order book (EOBI) multicast is provided to Co-location
members at no cost.

Service Type Details

Co-location Network Bandwidth 10 Gbps fiber optic network
10G Fibre Connectivity Available
Roundtrip Latency (in µsec): Exchange Response Time 6
Roundtrip Latency (in µsec): Colo Network* 10
Roundtrip Latency (in µsec): Exchange Response Time +
16*
Colo Network Latency

In keeping up with the global trends and in continuation of service excellence, NSE is facilitating
its members to co-locate their DMA and ALGO IT infrastructure at NSEIL premises. The co-
location facility would have the following features:
 Co-location facility shall be used only for DMA and Algo trading on NSE.
 Co-location facilitates with dual UPS power source and 100% DG capacity which ensures
uninterrupted power.

37
  Multiple Precision air conditioning units, with N+1 redundancy ensuring optimal temperature
at all times.
 Co-location facility at the Exchange comes in 2 variants: Standard 42U Full Rack with 6KVA
power and Half rack of 21U with 3.5KVA.

4.5 CO-LOCATION ARCHITECTURE

Diagrammatic representation of Sponsored Market Access under the Technology Programme of

BSE

Racks:
There are 3 types of racks provided by Netmagic. Members can apply for the rack type as per their
requirement.
1. Quarter Rack with 10 Gbps Fibre Link
2. Half Rack with 10 Gbps Fibre Link
3. Full Rack with 10 Gbps Fibre Link

Connectivity:
The members, who utilize the facilities of the colocation, will also be provided with the following
services to connect to their servers:-
1. Remote Secure Access (1 Mbps Internet connection with Firewall)
2. Additional 1 Mbps Capped Internet Bandwidth
3. Cross Connect for copper including Patch Cord

38
The Members can avail 'Sponsored Market Access' (SMA) at BSE Co-location and features
thereof are:

 SMA is a software distribution model in which applications are hosted by a service

provider and made available to customers over a network.
 Free of cost rack space is allocated by BSE to the SMA Service providers to install their
algo trading application server at Co-location. The members can connect to this SMA algo
server at Co-location via Internet.
 The algo strategy parameters can be triggered from the member front-end application and
sent to the SMA algo server at Co-location.
 The corresponding algos will generate orders to be sent to BSE trading system through
Co-location network.
 The algo orders will be sent to the BSE Trading system via the high speed Co-location
gateway for minimum latency.
 Members can execute orders over high speed Co-location network without installation of
any infrastructure at Co-location by the members.

10Gbps network at BSE Co-location:

 Low latency 10Gbps switches deployed at Co-location.

 Dedicated Co-location gateways with minimum hops to route the algo orders to BSE
Trading system for low round trip latency.
 Connectivity between members' Colo Racks and Co-location switches upgraded from 1
Gbps Copper to 10 Gbps Fibre network.
 Members' trading servers directly connected to Co-location low latency switches via 10
Gbps Fibre network for low latency trading.
 Low latency 10 Gbps Solarflare NIC cards (Model No. : SFN7322F) is provided to
members by BSE.

Advantages of SMA at BSE Co-location –

Advantages for Members -
 Zero capital expenditure on infrastructure at Co-location.
 The algo orders will be generated at Co-location and sent to the BSE Trading system via
the high speed 10G Co-location network for minimum latency.

Advantages for Vendors -

 Free of cost rack space is allocated by BSE to SMA Service providers to install their algo
trading application server in BSE Co-location.
 Vendor can service multiple clients with a centralized algo server installed at Co-location.
Hence, vendor's cost of maintenance is low.

39
4.6 QUANTUM OF ORDERS EMANATING FROM CO-LOCATION

Count of 25,155 million algorithmic orders generated in co-location servers across segments (CM,
F&O and CDS) for the period FY 16-17 (Apr’16 to Feb’17).

Table 4.1: The segment wise percentage of Algo orders coming from co-location

% COLO ORDERS
% ALGO ORDERS
OUT OF TOTAL
SEGMENT OUT OF TOTAL
ALGO ORDERS IN
ORDERS IN BSE
BSE

EQUITY 98.56 80.00

EQUITY DERIVATIVES 99.89 80.10

CURRENCY DERIVATIVES 94.93 91.88

INTEREST RATE DERIVATIVES 93.22 96.16

4.7 COLOCATION COSTS AT NSE

*(Charges in Rs. Lacs)

Rack Size Charges per rack p.a.( incl basic IT charges) *

Full Rack 12.00

Half Rack 6.00

Any taxes, duties, levies etc. would be extra, as applicable.

Key Points
a. Backup restore facility will be charged at Rs. 66000/- per rack. (Taxes, as applicable, would be
extra)
b. For a new applicant, a one-time initial set up charge of Rs. 1,00,000/- for a Full rack and Rs.
50,000/- for a Half rack would be levied. (Taxes, as applicable, would be extra)

Additional Costs apart from Rack Charges

Apart from rack charges, member may incur the following costs which varies from member to
member depending upon their choice:-
A) Cost for each interactive connection varies from Rs. 50,000 to Rs. 10,00,000 p.a. plus taxes
depending upon the message capacity of the connection
B) Cost of availing Tick By Tick data is Rs. 1,00,000 per connection p.a. plus taxes
C) Optional Tape Back Up charges of Rs. 66,000 p.a. plus taxes

40
Apart from the above Exchange’s costs, the trading members need to deploy requisite hardware,
software and connectivity of their choice at their cost.

Connectivity Costs
Members can choose from various connectivity options available for market data and order message
connectivity.

NSE uses TCP/IP protocol based Wide Area Network facilitating standard/higher bandwidth, expansion
and scalability.
In order to provide equal access to all the Trading Members spread over a wide geographical area, the
Exchange offers the following forms of telecommunication connectivity:
 VSAT (Very Small Aperture Terminal) - Satellite-based Connectivity
 Leased Line-Terrestrial-based Connectivity
 MPLS (Multiprotocol label switching)
Trading Members are required to choose a scenario from the available categories to apply for
connectivity.

Members may take one or more leased line to the co-location facility from different telecom service
providers for the purpose of setting up or modifying parameters, trading related activities and hardware,
software, network related access, software download / upload and monitoring and data downloads.

Table 4.2: VSAT, Leased Line and MPLS connectivity cost (in Rs.)
One Time
Deposit/Charges
Annual Charges
Interest
Free
Security Installation Hardware
Type of Deposit Charges Annual Recovery Last Mile Rent
Connectivity (IFSD) (approx) Charges(ARC) Charges (approx)
A) VSAT
1) Option I:
Premium - 17750 36000 - -
2) Option II:
Standard - 17750 23500 - -
B) Leased Line ( All bandwidths )
1) Category A
(40 m.p.s) 100000 50000
2) Category B
(100 m.p.s) 250000 250000
3) Category C
(200 m.p.s) 500000 500000
As per As per
4) Category D
As per service service service
(400 m.p.s) 1000000 1000000
providers rate providers providers
5) Category T (#) 100000 card 100000 rate card rate card

41
 6) Category S 100000 50000
C) MPLS
Category Category Category Category Category
Particulars A A B C D
Bandwidth 2Mbps 4 / 6 / 8 / 10 /20 Mbps
Interactive
messages per
second 40 40 100 200 400
Message Annual
Recovery Charges
(excluding service
tax) NA NA 2,50,000 5,00,000 10,00,000
Other Charges (Last
mile /Port Charges, Rate Charts of Service Provider as per NSE Circular Download Ref No.
etc.) NSE/MEM/31583 Dated January 18 , 2016

 The above charges are to be paid to the service provider except as follows:
i. In case of any leased lines emanating from POPs located at Rajkot, Jaipur and Cochin,
the IFSD and ARC shall be payable to NSE
ii. In case of category B, C & D Leased Lines, the IFSD shall be payable to NSE
Other than ARC, other annual charges may vary depending on the service provider
VSAT, MPLS and category S are not available at co-location
First category a connection on a Rack in co-location shall be provided free. Applicable ARC
shall be collected by NSE for connections at co-location
All above costs are exclusive of taxes. Applicable taxes will be charged extra.
(#) In case of availing Multicast TBT data on Higher Bandwidth Leased Lines, IFSD will not be
applicable. Applicable ARC shall be payable to NSE.

4.8 COLOCATION COSTS AT BSE

Exchange provides open access to all members to setup in the Exchange colocation. Trading members
do not have to incur any cost for setting up servers in BSE co-location. The Exchange provides both
rack space and servers to the members.

Technology Programme (Software and Hardware)

In order to facilitate usage of newer technology/development amongst the market participants, BSE
has launched a Technology Programme. The details are given below:

Under the Technology Programme, BSE bears the cost on behalf of the member for:
 The software license cost for Algo trading and market access. Algo software is provided to
members for trading from their Co-location and Non Co-location sites registered with Exchange.
 The infrastructure cost for Co-location rack space
 The Algo trading servers at Co-location
 The internet connectivity between BSE Co-location and member's office

42
 Table 4.3: Co-location Infrastructure Cost
Full Rack Free
Half Rack Free
Quarter Rack Free
Servers Free
Algo software Free
10G Switch at Member Rack Not Mandatory
Additional cost to member for 10G
NIL
migration
10G NIC Card from Exchange per
2 Solarflare Cards (SFN7322F) - Free
member
Interactive and Market data connectivity No cost for order throughput and market data
cost feed

Additional Costs apart from Rack Space: While there is no cost charged by BSE to members for
setting up colocation infrastructure, a member may have to incur their internal costs for development
and maintenance of their software application for ALGO trading, human resources, lease lines and any
other service that is not provided by BSE.

4.9 REQUIREMENTS TO BE MET FOR SETTING UP COLOCATION SERVER

Currently, to use colocation for trading purpose, Exchange approval for algo and/or DMA product is a
prerequisite.

A member needs to apply through BEFS (online member service portal of BSE) for the colocation
services required. This application is processed post approval from Business and Operations teams of
BSE.

4.10 LENGTH AND QUALITY OF WIRE USED

The colocation service is uniform for all members. We assume that the wire being discussed refers to
the connection of colocation racks with Exchange systems. All racks are equidistance with a maximum
tolerance of 0.43% and the tech specs of the connection is 50 Micron Lazerspeed OM3 MM Fiber.

Yes, the wire quality and length is same for all the members located at co-location. The wire length is
20 meters (from Colocation switch to Members’ servers in colocation racks) and the wire type is Cat6
Multimode Fiber optic.

43
4.11 SYSTEM (UNICAST/MULTICAST) IN EXISTENCE FOR THE LAST 5 YEARS

The market data (TBT) system was available in unicast (TCP/IP) till December 2016. At present only
multicast TBT is available which was introduced in May 2014.
A. TCP TBT (discontinued in December 2016) – The system was divided into two modules –
i. Primary source which receives data from trading system and converts it into TBT data
format
ii. Primary/secondary disseminator which accepts and maintains user sessions and sends
data to each connected used in round robin fashion.

44
 B. Multicast TBT – The data feed is disseminated directly from trading system over multicast
channel. In this system all users can listen to a predefined multicast IP address and port and
receive the data over multicast.

The market data is provided on multicast only. The exchange supports full order book dissemination in
its Enhanced Order book interface, EOBI. Another variant of this is the EMDI, Enhanced Market data
interface, which provides real time updates for top 5 price points, and netted data for other depths. For
trading terminals a low bandwidth netted data stream called MDI, market data interface, is also
supported. All these above mentioned streams are disseminated on 2 incremental channels for
redundancy supported by 2 redundant snapshot channels. All such channels are multicast based. The
exchange also supports its older market data protocol called NFCAST, which is a periodic snapshot
based market depth feed.

45
4.12 POSSIBILITY OF UNDUE ADVANTAGE TO SOME PARTICIPANTS DUE TO
SEQUENTIAL ACCESS DUE TO EARLY LOG-IN

Currently only multicast TBT data feed is available. The multicast protocol does not require any user
to login to a system. The data is sent over a multicast channel and network switches take care of sending
data to all the users who are connected and are receiving the data. The network switches disseminate
the data in parallel to all the users connected to it. Hence there is no undue advantage in multicast TBT
data feed.

No participant can gain undue advantage as the market data is multicast to all users and there is no
requirement for any login for getting market data.

For order and trades, the Gateways are assigned randomly based on load balancing logic built into the
session management process.

4.13 REQUIREMENT OF MAIN SERVER and BACK UP SERVER?

In multicast TBT, there is no concept of main and backup server. Currently, there are two multicast
feeds available - A and A’. Both the feeds are in active-active mode. The user can listen on either or
both the feeds simultaneously. Both the feeds receive data from same source and in same manner. There
are no additional charges for the data feeds.

Since all market data is multicast, there is no requirement of having a main or backup server. Due to
the multicast protocol, there is no scope of preferential treatment.

46
Chapter-5

ORDER TO TRADE RATIOS

47
 Chapter 5
ORDER TO TRADE RATIOS

5.1 INTRODUCTION
Order-to-trade (or order-to-execution) ratios involve financially penalising individual financial firms if
the orders to buy or sell they enter do not lead to a ‘sufficient’ number of trades.

High order-to-trade ratios imply that market participants are placing and cancelling orders but not
executing most of the orders. This could be due to the nature of market making algorithms or market
manipulation algorithms, where orders are placed to drive volumes to that point and then cancelled –
with the result that most of the orders are not converted into trades.

5.2 AVERAGE ORDER TO TRADE RATIO FOR ALL ACTIVE ALGO PARTICIPANTS IN THE
LAST 3 YEARS (YEAR-WISE BIFURCATION)
Order to trade ratio computed for all algo participants (trading members) on NSE irrespective of
colocation or non-colocation across segments (CM, F&O and CDS) for the last 3 financial years.
Average of all the participants (trading members) is provided below:

Table 5.1: Average Order to Trade Ratio for all active Algo Participants
FY Average of Order to Trade Ratio
2014-2015 8.91
2015-2016 26.44
2016-2017 (Apr’16 to Feb’17) 12.26

Notes: Algorithmic order is identified based on identification code as provided by the trading
member.

5.3 ORDER-TO-TRADE RATIO FOR THE TOP 10 PARTICIPANTS (BY TURNOVER) ACROSS
THE LAST 3 YEARS (YEAR-WISE BIFURCATION)

Order to trade ratio computed for all Algo participants (trading members) irrespective of colocation or
non-colocation across segments (CM, F&O and CDS) for the last 3 financial years. The top 10
participants (trading members) based on algorithmic trading turnover identified for the respective year
and their order to ratio for the respective year is provided below:

48
 Table 5.2: Order to Trade Ratio for top 10 Algo Participants (by turnover)
Participant Rank -
2016-2017
Based on Algo 2014-2015 2015-2016
(Apr’16 to Feb’17)
Turnover
1 6.43 6.19 7.99
2 7.50 14.94 8.68
3 0.74 2.24 31.62
4 4.30 9.28 2.53
5 2.62 0.77 2.77
6 3.91 1.76 7.81
7 2.04 10.96 29.41
8 1.75 32.78 2.89
9 59.50 28.16 1.77
10 8.69 3.97 16.98

Notes:
1. In case of options, premium turnover is considered in F&O and CD segment.
2. Algorithmic order is identified based on identification code as provided by the trading member.

5.4 ORDER-TO-TRADE RATIOS FOR EFFICIENT VS INEFFICIENT MEMBERS ACROSS THE

LAST 3 YEARS (YEAR-WISE BIFURCATION)

Order to trade ratio computed for all algo participants (trading members) irrespective of colocation or
non-colocation across segments (CM, F&O and CDS) for the last 3 financial years. The top 25
participants (Efficient members) and bottom 25 participants (Non Efficient Members) based on
algorithmic order to trade ratio identified for the respective year. The category wise average order to
trade ratio for these 25 members in the respective financial year is provided below:

Table 5.3: Average Order to Trade Ratio for Efficient and Non-Efficient Members (NSE)
FY Efficient Members Non Efficient Members
2014-2015 0.21 91.02
2015-2016 0.17 433.13
2016-2017 (Apr’16 to Feb’17) 0.14 181.47

Notes: Algorithmic order is identified based on identification code as provided by the trading member.

49
The below table provides the bifurcation for efficient vs non-efficient members (based on top and
bottom decile categorization) on the Equity segment at BSE:

Table 5.4: Average Order to Trade Ratio for Efficient and Non-Efficient Members (BSE)
FY Efficient Members Non Efficient Members
2014-2015 0.53 2230
2015-2016 0.85 4591
2016-2017 (Apr’16 to Feb’17) 0.85 1647

5.5 TREND OF ORDER-TO-TRADE RATIO OVER THE LAST 3 YEARS

Number of orders entered and executed into trades across all segments (CM, F&O and CDS) by all algo
participants (trading members) considered to arrive at the order to trade ratio for the respective financial
year.
FY Order to Trade Ratio
2014-2015 7.07
2015-2016 12.31
2016-2017 (Apr’16 to Feb’17) 11.22

The trend of order-to-trade ratio over the last 3 years (year wise – segment wise) is as given in the table
below:

Table 5.5: Year wise trend of order to trade ratio of ALGO trading members

SEGMENT YEAR

2016-17
(Till February
28,2017) 2015-16 2014-15

Avg. Order to Avg. Order to Avg. Order to

Trade Ratio Trade Ratio Trade Ratio

Equity 218.13 571.21 321.57

Equity Derivatives 3597.70 3083.59 8612.31

Currency Derivatives 195.22 290.42 111.09

Interest Rate
Derivatives 132.87 968.82 884.55

50
5.6 EXCHANGE PRO-ACTIVENESS IN CONTROLLING THE ORDER-TO-TRADE RATIO

Exchange on an Intra-day basis makes calls and alert trading members in case of High Order to Trade
ratio, these calls act as a preventive measure. From April 2016 till date Exchange has made about 110
calls which resulted into lesser disablements. Exchange also held member meet to make them aware of
the trends in the market which help them benchmark their order to trade ratio etc.

In Equity Segment the Order to trade ratio stood at 6.95 in November 2016 as against 9.41 in April
2016. Similarly in Equity derivatives segment the Order to trade ratio stood at 151 in November 2016
as against 215 in April 2016.

Further, Exchange also levies fair usage charges for multi leg order in derivatives segment such as 2
Leg and 3 Leg (2L/3L) orders which are Immediate or Cancel (IOC) in nature. Since this facility enables
trading members to enter orders in multiple contracts through single order entry, order to trade ratio
tends to be higher. In order to facilitate and encourage fair usage of such order entry facility, trading
members are charged for 2 Leg and 3 Leg orders in equity and currency derivatives segment.

Exchange has imposed throttles on the order flow to keep a check on trading members’ Algos as per
Exchange circular ref. 20161019-35 (Equity Segment), 20160930-3 (Currency derivative), 20160503-
22 (Equity Derivative.)

Exchange levies fair usage charges to trading members who do not satisfy the fair practices criteria
prescribed by the exchange as per Exchange circular no 20161027-14.

Circulars

20160930-3 - 20161019-35 - Order 20161027-14 - Fair 20160503-22 -Order

Currency Derivatives Order
flow rate
Flowfor
Rate.pdf
Equity.pdfUsage Charges.pdf Flow Rate for Equity Derivatives.pdf

51
5.7 PENALTIES OR STRUCTURES IN PLACE TO CURB HIGH ORDER-TO-TRADE RATIOS

Structure 1 (Penalty charges for higher order-to-trade ratios)

Table 5.6: Penalty charges levied at member level for high algo order to trade ratio (NSE)
Daily algo Order to Trade Ratio Charges (per algo order)
Less than 50 NIL
50 to less than 250 (on incremental basis) 2 paise
250 to less than 500 (on incremental basis) 10 paise
500 or more than 500 (on incremental basis)* 10 paise

*In case the ratio is 500 or more than 500 during a trading day, the concerned member is not permitted
to place any orders for the first 15 minutes on the next trading day(in the continuous trading session) as
a cooling off action.

Structure 2 (Fair usage charges categorized according to execution efficiency)

Further, charges are levied at member level for 2 Leg and 3 Leg orders. The charges are based on order
execution efficiency i.e. the ratio of number of trades resulting from 2L/3L orders and total number of
2L/3L orders as explained below:

Table 5.7: Fair usage charges categorized according to execution efficiency

Execution Efficiency % Levy on each 2L/3L order*
More than 90% NIL
Between 70% to 90% Rs. 0.01 (one paisa)
Between 50% to 70% Rs. 0.02 (two paisa)
Less than 50% Rs. 0.03 (three paisa)
*Applicable taxes extra

Execution Efficiency % = (Total number of trades executed by trading member / Total number of
Chargeable Algo orders) * 100

Structure 1 (Penalty charges for higher order-to-trade ratios)

Table 5.8: Penalty charges levied at member level for high algo order to trade ratio (BSE)

52
20130529-6 - Equity 20141010-26 - Order
Der Order to Trade ratio.pdf
to Trade ratio for Currency Derivative.pdf

Structure 2 (Fair usage charges categorized according to execution efficiency)

Table 5.9: Fair usage charges categorized according to execution efficiency

Slab Execution Efficiency % Fair Usage Charges per Chargeable Algo Order
* (in Rs)
1 Greater than or equal to 90% NIL

2 Greater than or equal to 70% but 0.0025

less than 90%

3 Greater than or equal to 50% but 0.0050

less than 70%

4 Less than 50% 0.0075

Following is the procedure to calculate Fair Usage charges:

1. Execution Efficiency % = (Total number of trades executed by trading member / Total

number of Chargeable Algo orders) * 100
2. For the computation of ‘Execution Efficiency %’,

Total number of trades executed by trading member:

a. All trades executed by a trading member resulting from the member’s algo and non-algo
orders across all instruments in a BSE trading segment shall be considered.

Total number of Chargeable Algo Order:

a. Chargeable Algo Order: This is defined as all active algo orders which are entered with a
price beyond the first 10 price levels or further than 1% of the best bid/ask order (whichever
is higher) prevailing at the time of entry in the order book of a given instrument shall be
considered. For this purpose, best bid shall be considered for the active buy algo order and
best ask for the active sell algo order.
b. Non-algo orders shall not be considered.
c. Algo order ‘add’ and ‘modify’ requests shall be considered. Order cancellation requests shall
not be considered.
d. Every modification of an algo order resulting in the order to be categorized as ‘Chargeable
Algo Order’ shall be considered as an additional algo order count
e. If an order is already a ‘Chargeable Algo Order’ then each modification of any parameter in
such order that continues to position it as a ‘Chargeable Algo Order’ shall be considered as
an additional algo order count.
f. If an algo order at the time of entry in the market is not a ‘Chargeable Algo Order’ but later
loses its price priority and is positioned beyond the first 10 price levels or further than 1%
of the best bid/ask order (whichever is higher) then it shall not be considered.
g. 2 lakhs ‘Chargeable Algo Orders’ per day per member in each trading segment will be
exempted.

53
 3. Charges shall be computed on a daily basis for all trading members. Sum of daily
charges shall be debited to the respective members at the end of a calendar month only
if the amount is equal to or greater than INR10,000 per trading segment for a particular
month.

5.8 PENALIZING REPEAT OFFENDERS OF HIGH ORDER-TO-TRADE RATIO

In order to discourage repetitive instances of high daily order-to-trade ratio, there is an additional
penalty in form of suspension of proprietary trading right of the trading member for the first trading
hour on the next trading day in case trading member is penalized for maintaining high daily order-to-
trade ratio, provided penalty was imposed on the trading member on more than ten occasions in the
previous thirty trading days.

In case of Trading members whose OTR is more than 500 in the derivatives market, such trading
members shall be placed in risk reduction mode for the first 15 minutes on the next trading day (i.e.
they can place orders only to reduce their existing position and not to increase their position) .

5.9 CAUSES OF A VERY HIGH ORDER-TO-TRADE RATIO

Based on our discussions with market participants, we are given to understand that order to trade ratios
are affected by various factors like type of strategy deployed, maturity/effectiveness of the algo as it
evolves, market conditions, volatility, liquidity etc. Further, market makers are known to use algos
extensively as they make markets in multitude of assets, scrips and contracts simultaneously which is
humanly not possible. It is further understood that the important function of market making involves
providing sitting quotes across large number of tradable books which may not necessarily result in
trades.

Algos with repeated order modifications are the primary cause of high OTRs.

5.10 WHAT CATEGORY OF HFT ALGORITHMS COULD LEAD TO HIGH ORDER-TO-

TRADE RATIOS

Trading members are not required to share the business logic of algo as it is a proprietary information.
Further market making activity may lead to high order to trade ratio.

HFT algorithms with high order modifications and low order execution resulting in trades generally
cause high Order to Trade ratio.

54
 Chapter-6

REGULATORY FRAMEWORK

55
 Chapter 6
REGULATORY FRAMWORK

6.1 OVERVIEW OF REGULATORY CONCERNS

Following are some of the regulatory concerns raised in the discussion paper by SEBI:

I. Algorithmic / high frequency trading has continued to attract the attention of investors and
regulators across the world during last few years. Some of such issues that have been
drawn regulatory attention are contribution to price volatility, market noise (excessive
order entry and cancellation), cost that high-frequency trading imposes on other market
users, technological arms race, limited opportunities for regulators to intervene during high
volatility, strengthening of surveillance mechanism, etc.

II. Fair, Transparent and Non-discriminatory access is one of the key pillars of a safe and
vibrant capital market. As some market participants across the globe have highlighted the
concern of unfair access and inequity to the non-colo / non-HFT participants vis-à-vis the
participants that use trading algorithms and co-location to trade, securities market
regulators are examining various proposals to address such concern.

6.2 MEASURES ADOPTED BY SECURITIES MARKETS REGULATORS IN DIFFERENT

COUNTRIES:

I. Minimum Resting Time

a. In June 2009, ICAP introduced a minimum quote lifespan (MQL) on its electronic
broking services (EBS) platform. These quote requirements set a minimum life of 250
milliseconds (ms) for their five ‘majors’ (generally currency contracts) and 1,500ms
in selected precious metals contracts. In public statements, ICAP credits the absence
of a major Flash Crash to MQLs

II. Frequent Batch Auctions

a. Taiwan Stock Exchange (TWSE) used to have continuous auction mechanism as the
order matching method wherein orders were batched over various time intervals.
TWSE has now moved to continuous limit order book mechanism for regular trading.
Auction methodology is used only for opening and closing price sessions.
b. Further, effective from April 2013, trading in illiquid stocks in the equity markets of
NSE, BSE, MSEI are conducted only through a periodic call auction mechanism.

III. Random Speed Bumps or Delays

a. ParFX, a wholesale electronic trading platform designed by Tradition (an interdealer

broker in over-the-counter financial and commodity-related products), applies
randomized pause to all order submissions, amendments and cancellations by between
20-80 milliseconds.
This limits the advantage of ‘first in, first out ‘trading and nullifies advantages gained
by low-latency trading strategies. It is understood that the objective is to provide a level

56
 playing field for participants wherever they are located and whatever their
technological or financial strength.
b. TSX Alpha Exchange (TSXA) imposes a randomized order processing delay of
between 1 and 3 milliseconds on all orders that have the potential to take liquidity. This
is intended to discourage opportunistic liquidity taking strategies. The intention is to
encourage orders to contribute to greater volume at the best bid/offer, translating to
larger trade sizes and better fill rates for active orders.
c. SEC (USA) has approved a proposal of IEX that non-routable Immediate-or-Cancel
(“IOC”) orders shall be subjected to a certain sub-millisecond delay before arriving at
the IEX system.

IV. Randomization of orders received during a period (1-2 seconds)

a. ICAP’s EBS Market Matching Platform has introduced ‘Latency Floor’ that consists
of a random batching window of 1, 2 or 3 milliseconds, whereby all messages
submitted within this period are collected and then randomly released to the matching
engine. The process is aimed at ensuring that speed as a stand-alone strategy is not a
pre-requisite for success on EBS Market.

V. Maximum order message to trade ratio requirement

a. The ICAP has a monthly fill ratio (MFR) requiring that at least 10% of all quotes
submitted into the market must result in an execution.
b. LSE’s Millennium trading system has message throttling constraints and penalties for
excessive ordering strategies. Anecdotal evidence suggests that the LSE message
policy was not fully effective in that it gave rise to new patterns of trade in low-priced
stocks. The LSE has experimented with changes in pricing effective May 4, 2010
whereby, among other measures, the threshold for the high usage surcharge for FTSE
350 securities increased from an OTR of 100/1 to a ratio of 500/1 (which is still the
figure in use at the time of writing). The frequency of order book updates nearly
doubled for a few months as a result before coming down again.

VI. Market Maker Pricing

a. Toronto Stock Exchange applied Market Maker Pricing :Examining the effects of a
controlled experiment on maker-taker pricing on the Toronto Stock Exchange,
Malinova and Park (2011) find that the bid-ask spread adjusted to reflect the breakdown
of maker-taker fees. They also found that the quoted depth of stocks eligible for maker-
taker pricing increased significantly, suggesting provision of greater liquidity.
Adjusting for the fees, the average bid-ask spread was the same before and after the
introduction of maker-taker pricing, and volume was greater for those stocks. Overall,
maker-taker fees improve markets by increasing depth and volume while holding
spreads (including fees) the same.
b. There was an introduction of maker-taker exchange fees for Australian securities cross-
listed on the New Zealand Stock Exchange in 2008. Berkman et al. (2011) found that
depth at the best quotes as well as trading activity increases with the introduction of
maker-taker fees, though there is little evidence of a change in bid-ask spread.

57
 c. It was introduced in European Markets. The only study focusing specifically on maker-
taker pricing in European markets identified by the Project is Lutat (2010). He finds
that the introduction of maker-taker pricing by the SWX Europe Exchange did not
affect spreads but led to an increase in the number of orders at the top of the book.
d. Dutch Index introduced Maker Taker Model. Menkveld (2012) finds that spreads fell
dramatically when Chi-X began trading Dutch index stocks, suggesting that its maker-
taker model may have improved market competitiveness

VII. HFT Transaction Tax

a. HFT transaction tax in France

In 2012, France introduced a national levy on certain financial transactions with French
shares and certain derivatives as well as cancelled orders in the context of ’high
frequency-trading’. In particular, the tax on cancelled (or modified) HFT orders (0.01%
of the value of the orders) applies in case the ratio of cancelled orders to all orders
during one trading day exceeds 80%. The tax is payable by each financial intermediary
(except if undertaken in the context of market making) established in France and using
automated algorithm trading characterized by the successive sending of purchase or
selling orders and the modification or cancellation of the initial trading orders,
respectively, within a time period of no more than half a second.

A subsequent report by the European Commission to assess the impact of the tax on
trading volumes, price levels and volatility concluded that while no significant impact
was seen on the volatility and price level of the stocks, a decline of 10% in liquidity
was observed.

b. HFT transaction tax in Italy

In 2013, Italy levied a 0.02% tax on order changes and cancellations occurring within
a time frame shorter than 0.5 seconds. The tax is calculated on a daily basis and is
payable where - in a single trading day, the ratio between the sums of cancelled and
modified orders, and the sum of entered and modified orders exceeds 60% with
reference to a single financial instrument. The tax is applicable from March 2013 for
equities and from July 2013 for derivatives.

58
VIII. Global regulatory response to algorithmic trading

a. The U.S. Securities and Exchange Commission (SEC)

The SEC has undertaken a series of steps overtime to prevent instability in the market
that may arise from automated trading. The most significant of these are: Market
Access Rule, Regulation SCI and registration of entities. These are:

1. Market access rule

In November 2010, the SEC approved a new rule to require brokers and dealers to have
risk controls in place before providing their customers with direct access to the market.
The rule prohibits broker-dealers from providing customers access with “unfiltered”
and “naked” access to an exchange or alternative trading system. Specifically, the rule
requires broker-dealers to:

 Create risk controls to prevent the entry of orders that exceed appropriate pre-set
credit or capital thresholds, or that appear to be erroneous.
 Create risk controls designed to ensure compliance with all regulatory
requirements applicable in connection with market access.
 Have certain risk management controls applied automatically on a pre-trade basis
before orders route to an exchange or ATS
 Maintain risk management controls and supervisory procedures under the direct
and exclusive control of the broker-dealer with market access except in limited
instances.
 Establish, document and maintain a system for regularly reviewing the
effectiveness of its risk management controls and for promptly addressing any
issues.
 SEC directed self-regulatory organizations to maintain a consolidated audit trail
(CAT). This audit trail is intended to increase the data available to regulators
investigating illegal activities such as insider trading and market manipulation.

2. Registration of persons responsible for algorithmic trading strategies

In April 2016, the SEC approved a rule which require registration of Securities Traders
of associated persons primarily responsible for the design, development or significant
modification of algorithmic trading strategies, or who are responsible for the day-today
supervision or direction of such activities.

3. Initiatives planned as per SEC Director’s testimony

Further initiatives planned as per SEC Director Stephen Luparello (Testimony given
on 3 March 2016):
a. Additional timestamps for consolidated data feeds and data feed usage by
Self-Regulatory Organizations or SROs.
b. Membership requirements for dealers trading in off-exchange venues.
c. Rules to strengthen record-keeping requirements for algorithmic trading.
d. Work on recommendations to clarify the status of active proprietary traders
not registered as broker-dealers.

59
 e. Work on recommendations to address the use of aggressive, destabilizing
trading strategies in conditions when they could most seriously exacerbate
price volatility.
f. In addition, the Equity Market Structure Advisory Committee has formed a
market quality subcommittee to consider the impact of technology on the
efficiency of the markets and systemic risk.

b. The German HFT Act

In May 2013, Germany enacted the Act on the Prevention of Risks and Abuse in High-
Frequency Trading (HFT Act).
i. Licensing: The HFT Act requires that firms engaged in HFT must be licensed. HFT
is defined to include each of the following elements:
o Trading for one’s own account, or by proprietary trading firms;
o Trading algorithmically without human intervention;
o Trading using low-latency infrastructures; and
o Trading that generates a high intra-day message rate.
ii. Risk Controls: Firms that engage in HFT-type business must fulfil requirements to
ensure that markets are not distorted or interrupted.
iii. Market Manipulation: The Act broadens the definition of market abuse.
iv. Order to Trade Ratio: The Act requires exchanges to impose, on a product-by-
product basis, an excessive system usage fee and an order-to-trade ratio limit
intended to prevent unnecessary messaging. Electronic Identification of
Algorithmic Trading: The Act mandates that ex-changes have to implement rules
requiring all exchange members to flag all algorithmically generated orders with a
unique key when sent to a German exchange so as to allow the market surveillance
system to allocate all orders to the generating algo.

c. Australia

In 2013, the Australian Securities and Investments Commission (ASIC) released its
consultation paper with regard to market integrity rules in the presence of HFT to
address: excessive messaging and market noise (in particular, small and fleeting orders
and order-to-trade ratios). The proposals that were open for public consultation
included:
i. Prevent small orders* being cancelled or amended within 500 milliseconds of
being submitted to the trading platform of a lit exchange market; and
ii. Establish systems, policies and procedures to prevent the cancellation or
amendment of small orders within 500 milliseconds of being submitted to the
trading platform of a lit exchange market.
*(For the purposes of the above proposal, ASIC proposed to define ‘small
order’ as being less than or equal to:
 $500 value for equities traded on the ASX and Chi-X markets;
 $500 value for contracts for difference traded on the ASX 24 market
 10 futures contracts for the ASX 24 market (for all contracts with the
exception of the ASX SPI 200 Index Future (ASX SPI 200 Future);
 Three futures contracts for the ASX SPI 200 Future)

60
 However, ASIC said that it decided to not go ahead with proposals to rest small orders
on the market for a set time or for dark orders to meet a minimum size.
At present, ASIC does not have any specific regulation on AT/HFT. Instead, it periodically
publishes a review of HFT and dark-liquidity in the Australian markets.

IX. Other Measures Employed:

a. In a bid to curb high-frequency trading, Australia has unveiled plans for the introduction
of 'kill switches' as part of a wider package of new market integrity rules. By the middle
of 2014, traders were required to have direct control over algorithms, including 'kill
switches' to immediately stop one if required and avoid a repeat of the infamous May 2010
US flash crash.

NSE has also employed kill switches as part of market integrity rules. However, the kill
switch is manual and the control given to the member for the execution of the same. Until
now, not even once has a kill switch been employed on NSE trades.

b. On April 15, 2014, the European Parliament approved new legislation to curb high
frequency trading by restricting the smallest price increments for securities and
requiring testing of trading algorithms to identify any behaviour mimicking front
running. However in EU, proposals to ensure orders stay on a trader's book for a
minimum time, making it difficult for high-frequency traders to operate, were dropped last
year.

c. Another measure introduced by EBS was to widen the spread of pricing on some of its
currency pairs, so the fifth decimal point in a quote was a “5” or “0” rather than increments
of a 10th. It also raised the ratio of quotes traders were allowed to send before they were
required to buy. The changes resulted in fewer “flash” orders – automatic trades made at
lightning speeds – at the top of the order book while trades further down the order book
were being filled

6.3 WHAT DOES IOSCO PRESCRIBE

Regulatory Issues Raised by the Impact of Technological Changes on Market Integrity and
Efficiency – Final Report by TECHNICAL COMMITTEE OF THE INTERNATIONAL
ORGANIZATION OF SECURITIES COMMISSIONS, October 2011.

IOSCO Recommendations on Market Integrity and Efficiency

i. Recommendations regarding trading venue operators and trading participants

a. Recommendations regarding trading venue operators

Recommendation 1: Regulators should require that trading venue operators provide

fair, transparent and non-discriminatory access to their markets and to associated
products and services.

Recommendation 2: Regulators should seek to ensure that trading venues have in place
suitable trading control mechanisms (such as trading halts, volatility interruptions,

61
 limit-up-limit-down controls, etc.) to deal with volatile market conditions. Trading
systems and algorithms should be robust and flexible such that they are capable of
dealing with, and adjusting to, evolving market conditions. In the case of trading
systems, this should include the ability to adjust to changes (including sudden
increases) in message traffic

b. Recommendations regarding trading participants

Recommendation 3: All order flow of trading participants, irrespective of whether they

are direct venue members or otherwise, must be subject to appropriate controls,
including automated pre-trade controls. These controls should be subject to the
regulatory requirements of a suitable market authority or authorities. In addition,
regulators should identify any risks arising from currently unregulated direct
members/participants of trading venues and, where any are identified, take concrete
steps to address them.

ii. Recommendations for regulators

a. Regulator’s knowledge of the markets and trading, and associated risk mitigation

Recommendation 4: Regulators should continue to assess the impact on market

integrity and efficiency of technological developments and market structure changes,
including algorithmic and high frequency trading. Based on this, regulators should seek
to ensure that suitable measures are taken to mitigate any related risks to market
integrity and efficiency, including any risks to price formation or to the resiliency and
stability of markets, to which such developments give rise.

b. Market surveillance and market abuse

Recommendation 5: Market authorities should monitor for novel forms or variations of

market abuse that may arise as a result of technological developments and take action
as necessary. They should also review their arrangements (including cross-border
information sharing arrangements) and capabilities for the continuous monitoring of
trading (including transactions, orders entered or orders cancelled) to help ensure that
they remain effective.

IOSCO’s panel sessions gave rise to numerous suggestions on how regulators’ tools could be
enhanced to address the risks associated with the technological and market developments that
have been seen over the past number of years.

Below are the possible tools to address Risk:

i. Level: Trading Firms:

a. Given that relatively few jurisdictions currently have regulations that are designed
specifically to address algorithmic trading or HFT, market authorities should consider
whether tailored regulatory requirements should be introduced, especially in those markets
where algorithmic trading or HFT is a dominant component of the market structure. Some
presenters suggested that this might include anything from specific stress testing and sign-

62
 off processes for new algorithms to specific charges or a tax on high order entry or
cancellation rates;

b. Consider whether those firms that are non-intermediary members of trading venues should
be subject to registration/authorisation by a regulator in those jurisdictions where this is not
already the case;

c. Reassess whether requirements for managing conflicts of interest are sufficient in the
circumstances where either:
 an investment firm simultaneously conducts client-serving activities and proprietary
trading; or
 trading participants that trade on venues in which they hold an ownership stake;

d. Review existing regulatory requirements regarding pre-trade risk controls applicable to

intermediaries in order to evaluate whether they are suitable for today’s high-speed
markets. As part of this, consider banning the provision of DEA to customers if the
customer‟s trading is not subject to appropriate pre-trade controls (i.e. so-called naked
access)

e. Assess whether HFT or algorithm traders should provide for specific forms of stress testing
and internal sign-off processes for new algorithms

ii. Level: Markets/Market Operators

a. Consider whether trading control mechanisms such as order entry controls to mitigate
against anomalous order entry and circuit breakers or limit-up/limit-down systems should
be mandated within the markets and, if so, whether venue operators should be permitted to
design their own controls or whether they should be harmonised/coordinated across venues
(including between interrelated instruments such as a derivative and its underlying).

b. Consider whether there should be common trade cancellation arrangements in place across
markets to ensure consistent treatment in the event of a sudden extreme price movement.
These arrangements should be coherent in their operation with any order entry controls and
volatility controls.

c. Consider requiring that market infrastructure operators undertake appropriate stress testing
to ensure that their systems are robust in the face of unusual spikes in trading activity

d. Consider whether to require market operators to have appropriate testing environments in

place to enable participants to stress test their algorithms;

e. Assess whether a trading venues registered market makers should be subject to mandatory
minimum criteria so as to ensure that they provide meaningful liquidity support to the
market. As part of this, consider:

 clarifying how market making should be defined; and

63
  banning so-called stub quotes (i.e. automatically-entered quotes that involve an
extremely low bid price e.g. 1c, and an extremely high offer price (e.g. $100k));

f. Assess whether specific charges, fees or taxes on high order entry and cancellation rates or
messaging rates should be introduced; and

g. Consider the introduction of minimum tick sizes and minimum order book resting time.

iii. Level: Market Structure

a. Reconsider the appropriate balance between encouraging competition between trading

spaces (including between trading venues and OTC space), which helps to reduce trading
fees and improves innovation, and promoting the use of transparent, on-venue trading;

b. Consider banning flash orders, through which trading interest may be exposed on a market
for less than a second during which only technically-adept participants are able to react to
it, before it is routed elsewhere;

c. Evaluate what could be done to improve market surveillance, taking into account the needs
of different market structures. A high-level of surveillance of potential unfair activity by
market participants is crucial. Up-to-date order screening/monitoring tools should be
implemented (either by trading venue operators or by competent authorities) to help
identify trading patterns and prevent inappropriate trading behaviour. Other possible
measures could include the introduction of consolidated “audit trails” that are able to track
orders, quotes and trades in the market. Other possibilities include introducing large trader
reporting requirements (where these do not already exist), and introducing the use of entity
identifiers to identify trading on a participant-by-participant basis or to flag
algorithmic/HFT orders; and

d. Review how existing market manipulation rules and laws apply to computer generated
orders and whether activity traditionally deemed manipulative is still appropriate in today‟s
market environment (e.g. layering the book given the common HFT strategy of submitting
orders at multiple price points).

Source:

3.-IOSCO-on-HFT-20 IOSCONEWS215.pdf
-October-2011.pdf

64
 Chapter-7

SEBI DISCUSSTION PAPER:

Strengthening of the Regulatory Framework
for Algorithmic Trading & Co-location

65
 SEBI DISCUSSTION PAPER:
Chapter 7 Strengthening of the Regulatory
Framework for Algorithmic Trading & Co-location

Below points have been raised in SEBI’s discussion paper and have been discussed in detail below
with pros and cons, global review of similar measures and NSE and BSE response to the discussion
points

7.1 MINIMUM RESTING TIME FOR ORDERS

i. Introduction

(a) Resting time is defined as the time between an order is received by the exchange and the said
order is allowed to be amended or cancelled thereafter.

(b) The proponents of algorithmic trading have always argued that it has improved liquidity and
depth of orders. The opponents of algorithmic trading have contended that the liquidity and
depth provided by trading algorithms is ‘Apparent’ and ‘Fleeting’ as it vanishes as the traders
intend to execute trade.

(c) This issue of ‘fleeting’ or ‘vanishing’ liquidity arises from the ability of the trading algorithms
to react to new developments (such as receipt of new order or market news) by usually
modifying / cancelling their orders or placing new orders. It is also gathered that such ability to
modify their orders has raised concerns with a section of market participants who consider that
this ability is prone to market abuse.

(d) In view of the above, securities market regulators / stock exchanges are considering / have
considered the idea to eliminate “fleeting orders” or orders that appear and then disappear
within a short period of time. As per the Minimum Resting Time mechanism, the orders
received by the stock exchange would not be allowed to be amended or cancelled before a
specified amount of time viz. 500 milliseconds is elapsed.

ii. Benefits

a. Can increase the likelihood of a viewed quote being available to trade

This has two important benefits:

 First, it provides the market with a better estimate of the current market price, something
which ‘flickering quotes’ caused by excessive order cancellations obfuscates.

 Secondly, its visible depth at the front of the book should be more aligned with the actual
depth. This knowledge of the depth improves the ability of traders to gauge the price impact
of potential trades. Quotes left further away from the current best bid or offer are less likely
to be affected by the measure since the likelihood of them being executed within a short
time is small. Nonetheless, minimum resting times might be expected to make the order
book dynamics more transparent to the market.

66
 b. Reduces the excessive level of message traffic currently found in electronic markets
Minimum resting times may also reduce the excessive level of message traffic currently
found in electronic markets. Cancellations and resubmissions are a large portion of these
messages, and at peak times they can overwhelm the technological capabilities of markets
(as seen for example in the recent Facebook initial public offering (IPO) problems on
NASDAQ) ( http://www.nanex.net/aqck/3099.html).

c. Allay concerns that markets are currently ‘unfair’ in that high frequency traders are
able to dominate trading by operating at speeds unavailable to other traders.
This notion of ‘slowing down’ markets is not generally supported by economic analyses,
but it does speak to the challenge of inducing participation if some traders, particularly
small retail investors, feel that speed makes markets unfair.

iii. Cost and Risks

a. Picking off stale orders

Preventing passive orders from being cancelled means that such orders are in constant
danger of becoming stale. This rule change will artificially increase adverse selection
effects for passive orders and will therefore cause an increase in the bid-ask spread as
passive orders will need to be compensated for this additional adverse selection risk.

What is Adverse Selection?

Suppose that in between the time when a passive sell order is placed and when it can be
cancelled, there is a positive news event that makes most market participants believe the
price should be higher. The passive order is unable to respond to this event, thus it becomes
a target to be picked off by an aggressive market order that buys the shares in the passive
order and then immediately sells them at a higher price. The aggressive market order thus
profits at the expense of the passive order. Such an effect is called adverse selection.

b. Raising transaction costs

Aggressive orders remove passive orders from the book, which means less liquidity.
Similarly, fewer passive order placements mean less liquidity. The net result is to increase
the bid-ask spread, which increases the cost to make a round-trip transaction. In general,
aggressive orders have more market impact, i.e. they cause larger changes in price. Thus
transaction costs will likely go up across the board, for both large and small orders, and for
both passive and aggressive orders.

c. Will hurt retail investors because of increased bid ask spread

It will cause an increase in bid-ask spreads and therefore increased costs to retail investors
not involved in the high frequency trading zero-sum game. Wider bid-ask spreads will also
mean that less trades happen as they impose a larger cost to trading that needs to be
exceeded by the benefits of each voluntary trade. Therefore the gains-from-trade will be
reduced.

d. Increased volatility
Less liquidity in the book may result in increased volatility. The price changes are noisy.
Under this view, decreased liquidity leads to larger price changes when an aggressive order
of a given size hits the market, thus causing more volatility.

67
 e. It is unfair to impose a regulation that will so clearly benefit one type of HFT
technology (aggressive) at the expense of another (passive)
Firms specialize and this will lead to a shuffling in the market ecology, with possibly
unpredictable consequences. Worse still, many market participants feel that it is fast
predatory market orders that cause unstable or volatile price moves, which may be an
additional reason to expect increases in volatility.

f. Change the dynamics of the market by attracting more aggressive high frequency
traders whose sole aim is to take advantage of the free options
Depending on the length of compulsory resting, those limit orders close to the best bid or
offer are likely to become stale (that is, no longer at the efficient price) before they can be
cancelled. This can spawn ‘front running’ by automated traders who collect the low-
hanging fruit from such options. In return, the providers of passive quotes will protect them
against staleness through yet larger bid-ask spreads, or by simply not posting quotes at all.
Using the estimates by Farmer and Skouras, the cost of hitting such stale quotes may be as
high as €1.33 billion per year in Europe.

g. High frequency traders may reduce their market making activities and possibly be
replaced by institutional market makers
Reduced competition among market makers and their need to earn a return on their capital
may also drive up transaction costs for end users. Moreover, to the extent that minimum
resting times inhibit arbitrage between markets, which is essentially at the heart of many
HFT strategies, the efficiency of price determination may be diminished

h. Reduced liquidity around volatile periods

During periods of high market volatility, new information enters more frequently than
during normal periods. Around such periods, quoted prices can become stale in very short
time intervals, based on the new information. Without the ability to modify / cancel the
quotes, market participants would be reluctant to place their orders when it is most needed.
This will make market liquidity less resilient and more un-stable around high information
periods due to the imposition of minimum resting time.

i. HFT firms in the future may just redirect their gaze to pinpoint 499 and 501
milliseconds

This can be done to generate their profits by picking off traders whose orders have remained
in clear sight for half a second or more.

iv. Ambiguity

It is unclear if IOC orders will also be subject to the minimum resting time. If such orders
continue to remain available, then there will be a movement from placement of limit orders to
IOC orders by participants. This will not be an optimal solution as this will reduce the depth in
the markets. If the restriction of minimum resting time is also imposed on IOC orders, these
orders will lose their importance.

68
v. Evidence from Global Markets

a. Currently, there are no instances of the ‘resting time’ mechanism being mandated by any
regulator. It has been observed that Australian Securities and Investment Commission (ASIC)
had sought feedback on the matter few years ago, but decided not to go ahead with the
proposal.

b. The commissioned study EIA3, examined the effects of minimum resting times inside
simulated market. They did not recommend its adoption.

c. In June 2009, ICAP introduced a minimum quote lifespan (MQL) on its electronic broking
services (EBS) platform. These quote requirements set a minimum life of 250 milliseconds
(ms) for their five ‘majors’ (generally currency contracts) and 1,500ms in selected precious
metals contracts. In public statements, ICAP credits the absence of a major Flash Crash to
MQLs

vi. Evidence from Indian Markets

As per IGIDR Comments on “SEBI’s Discussion Paper Strengthening of the Regulatory
framework for Algorithmic Trading &Co-location” :using data for NSE spot and single stock
futures (SSF) segment, the proportion and placement of orders cancelled in less than one second
was analyzed. The analysis is done for the 150 derivative securities traded on NSE for
November – December 2013.

Key Observations are:

 Spot market experiences lower percentage of cancellations in one second than the SSF
market (36.83% in Q1 stocks on spot versus 70.05% on SSF).
 Across all stock quartiles, less than 8% of the orders are cancelled within one second while
they are at best prices.
 A majority of the cancellations that occur in less than one second of arrival take place when
the order is away from top five prices in the order book.
The above observations suggest that that less than 8% of the orders could be called fleeting
orders. The remaining cancellations in less than one second occur when the order is far
away from the touch.

vii. NSE’s view is that resting time mechanism may be counterproductive and may lead to more
order being pumped into the system and longer queues and waiting time, besides complexity
induced in trading software.

69
viii. BSE’s view

Objective
 To prevent repeated actions on orders and eliminate fleeting orders.
 Matching engine will validate timestamp of book order with incoming
modification\cancellation request and accept only if it is compliance with resting
period, else the request will be rejected.
 The matching engine will perform this additional validation in addition to the matching
process itself

Comments

 More stability in limit order book

 Reduce fleeting orders
 Increased possibility of orders getting matched as they will be retained in the order
book

7.2 FREQUENT BATCH AUCTIONS (PERIODIC CALL AUCTIONS)

i. Introduction

a) Under the ‘continuous matching’ system deployed by the stock exchanges, the buy and
sell orders received by an exchange are continuously matched and resultant trades take
place.

b) The mechanism of Frequent Batch Auctions would accumulate buy and sell orders on
the order book for a particular length of time (say 100 milliseconds). At the end of
every such period, the exchange would match orders received during the time interval.

c) This proposal tries to address the problem of ‘latency advantage’ by undertaking batch
auctions at a particular interval. The idea is to set a time interval for matching of orders
which is short enough to allow for opportunities for intraday price discovery, but long
enough to minimize the latency advantage of HFT to a large extent.

d) For example: if one-way observed latency for a co-located participant at an exchange

and an investor located at New Delhi is 1ms and 15ms respectively, Batch Auctions at
every 20ms - 30ms may offer a fair chance to non-colocated participants to capture a
trading opportunity.

e) Expected impact: The proposal may nullify the latency advantage of the co-located
players to a large extent. However, due to batch auction sessions happening every few
milliseconds, the market infrastructure may require corresponding changes.

70
 ii. Ambiguity

It is not clear whether the proposal is meant for all segments of equity markets, or any one particular
segment. It is also not clear what time interval will be used for the proposed batch auctions. What
information will be supplied during the batch auction period is also not clear. Whether only the
equilibrium price will be indicated or whether the top five prices will be indicated is not clear. Further,
it’s also not clear if the orders will be executed on pro-rata basis, or on time priority basis.

iii. Benefits

a. Reduction of the speed of trading and the elimination of the arms race for speed
The speed of trading could be controlled through the timing and frequency parameters which
could be tuned to individual and market conditions. It might increase liquidity or at least
concentrate it at particular time points.

b. Eliminate Sniping
The proposal draws from Budish et al Research Paper, which argues that frequent
batch auctions could eliminate sniping. The paper also claims that the measure can
reduce the potential for manipulative activities such as spoofing and layering.

iv. Costs and Risks

a. Increased execution risk

Investors will face this risk if they do not know when the auctions will take place and then
whether their bid will be in the winning set. If the frequency of auctions does not reflect
market conditions, there may be some shortfall in matching supply and demand.

b. May reduce Liquidity

Since this is a proposal and has not been tested anywhere in the world, it is unclear how
frequent batch auctions will affect liquidity. In the presence of one exchange offering
continuous trading (example: SGX for Nifty futures) and another offering batch auctions
(NSE for Nifty futures), traders will prefer immediacy and migrate to the SGX. This will
exacerbate the fragmentation of a single market for the security, and can drive away
liquidity from domestic markets.

c. Reduce incentives to supply liquidity

Periodic call auctions would have a severe impact on the business model of market makers
and may reduce incentives to supply liquidity

d. Lead to drastic change in Trading Landscape

The current market structure allows a variety of different trading mechanisms and allows
for auction-based trading as well as the continuous limit order book (in both lit and dark
modes in international markets). This proposal would limit choice and require quite a
drastic change in the trading landscape. Furthermore, it would require coordination at the
global level, which may be difficult to achieve.

71
 e. Cross-market arbitrage
The implementation will require perfect time synchronicity across exchanges for cross-
market arbitrage. If not, then it will raise the arbitrage risk and can adversely affect the
pricing efficiency across two markets auctions.

f. Increased adverse selection costs

The implementation of batch auctions could increase adverse selection costs for traders by
making the quotes stale as in the case of an intervention that forces a minimum resting time.
This will especially be true of high volatility periods when information arrival in the market
is high

Evidence: Haas and Zoican, (2016) “HFT Competition and Liquidity on Batch Auction
Markets”

g. Reduced transparency
In a continuous market mechanism, a trader sending a market order can compute execution
price for his trades. Such markets update prices in real terms and provide price discovery
in its true sense. This will not be possible if markets move to frequent batch auctions model

h. Implementation costs
The implementation of frequent batch auctions requires incurring the cost of a structural
change in the market infrastructure from continuous trading. This means that all current
market institutions and processes (including the calculation of margins for clearing and risk
management such as price bands and margin limits). This cost will be imposed on all market
participants which makes this an expensive approach.

v. Evidence in Global Markets

Taiwan Stock Exchange (TWSE) used to have continuous auction mechanism as the order
matching method wherein orders were batched over various time intervals. TWSE has now
moved to continuous limit order book mechanism for regular trading. Auction
methodology is used only for opening and closing price sessions.

Cheng and Kang (2007) analyzed the impact of the new mechanism on market quality of
TAIFEX using intra-day data. They find that, “the market is more liquid, and volatility is
slightly lower, under continuous auction than under call auction. Also, there is robust
evidence that continuous auction improves informative efficiency.”

vi. Evidence from Indian Markets

Trading in illiquid stocks in the equity markets of NSE, BSE, MSEI are conducted only
through a periodic call auction mechanism from April 2013

i. As per NSE, Call auction mechanism introduced in Indian markets have had negative
impact in terms of:-

i. Liquidity in those securities

72
 ii. Spreads have increased

iii. Overall investor’s participation has been adversely impacted.

ii. NSE’s view is that introduction of call auction market would be an extremely retrograde
step for our markets and seriously impede price discovery. No large and progressive
market in the world users an auction for regular trading session.

iii. BSE’s Views

Objective
a. Replace continuous matching by multiple and frequent call auctions.
b. This is a change in the market scheduler, no code change is required.
c. Risk is of one auction overflowing into another.

Observations
a. Fundamentally change the core matching model from continuous matching to a
staggered matching.

7.3 RANDOM SPEED BUMPS OR DELAYS IN ORDER PROCESSING / MATCHING

i. Introduction

The Speed Bump mechanism involves introduction of randomized order processing delay of few
milliseconds to orders.

ii. Benefits

a. Nullify Latency Advantage

The intent behind such mechanism is to nullify the latency advantage of the co-located players
to a large extent. The delays would affect HFT but would not deter non-algo order flow for
which delay in milliseconds is insignificant.

b. Could stop the arms race for speed

However it is not clear how much it will nullify the speed advantage of HFT. No positive impact
on any other HFT related concern is expected.

iii. Costs and Risks

1. Increase Uncertainty and adverse selection costs

The randomness element will add to the uncertainty on when the order will go to the exchange,
and will increase the adverse selection costs.

73
 2. Reduced / withdrawal of liquidity
The reduced participation because of higher adverse selection costs will likely be exacerbated
during high information periods and reduce liquidity to traders and investors when it is needed
most. Finally, for products which are traded on international markets (such as the SGX, DGCX),
such a measure is likely to drive away liquidity from domestic markets. This will lead to higher
price discovery of Indian assets on foreign shores.

3. Increased price disparities across different market segments

Since this intervention increases uncertainty for traders, and reduced participation, it would
increase persistence of violation of cross-market arbitrage and worsen price inefficiencies across
market segments.

4. Implementation costs
The implementation of speed bumps will a direct cost for the exchange, and will raise costs of
trading to end investors.

iv. Evidence from Global Markets

a. ParFX, a wholesale electronic trading platform designed by Tradition (an interdealer

broker in over-the-counter financial and commodity-related products), applies
randomized pause to all order submissions, amendments and cancellations by between
20-80 milliseconds. This limits the advantage of ‘first in, first out’ trading and nullifies
advantages gained by low-latency trading strategies. It is understood that the objective is
to provide a level playing field for participants wherever they are located and whatever
their technological or financial strength.

b. TSX Alpha Exchange (TSXA) imposes a randomized order processing delay of between
1 and 3 milliseconds on all orders that have the potential to take liquidity. This is intended
to discourage opportunistic liquidity taking strategies. The intention is to encourage
orders to contribute to greater volume at the best bid/offer, translating to larger trade sizes
and better fill rates for active orders.

c. SEC (USA) has approved a proposal of IEX that non-routable Immediate-or-Cancel

(“IOC”) orders shall be subjected to a certain sub-millisecond delay before arriving at the
IEX system.

d. As per Thomson Reuters, it will be introducing a mechanism for its FX Spot Matching
services that introduces a short delay of several milliseconds before processing orders.

v. Evidence from Indian markets

As per IGIDR Comments on “SEBI’s Discussion Paper Strengthening of the Regulatory

framework for Algorithmic Trading &Co-location” the average time to modifications of an
order by algorithmic traders and non-algorithmic traders for Nifty 50 and Nifty Next set of
stocks was calculated. The analysis is done for the NSE spot market for the period of 2013.

74
 The median time between modifications of algorithmic traders is recorded as 0.21 second on
Nifty stocks. The same for non-algorithmic traders is 97 seconds. This indicates a significant
difference between the reaction time of algorithmic and non-algorithmic traders. However, this
doesn’t come as a surprise since algorithmic traders are expected to be much faster than the
non-algorithmic traders. (Find Table below for reference)

Given the difference, the question that arises is: will the speed bump of some milliseconds bring
the latency experienced by algorithmic traders closer to the latency of non-algorithmic traders,
and will it make the market a level playing field? The answer to the question appears to be a
no.

Average time between modifications on NSE spot segment:

The table below presents average time between modifications for Nifty 50 and Nifty Next
stocks on the NSE spot market in 2013. Algo represents average time between modifications
in an order by algorithmic traders, while non-algo represents average time between
modifications by non-algorithmic traders.

Table 7.1: Average Time between Modification for Nifty 50 and Nifty Next Stocks o the NSE
Spot Market (2013)

All values in milliseconds

Nifty stocks Nifty Next stocks

Algo Non algo Algo Non algo

Q1 26 6,847 27 8,366

Median 211 97,220 371 89,943

Q3 1,400 1,638,676 2,252 1,548,837

vi. In NSE view, speed bumps will not deter algo traders as it can be programmed to cushion
against the bumps.

vii. BSE’s view

Objective
a. Forcefully delay orders with random time delays before taking up for matching. Orders
will continue to have price time priority, with a random delay introduced in the flow.

75
 b. Randomization logic is of utmost importance. Orders will be scheduled for processing
after a random delay within a specified time range and taken up for matching after the
delay.

Comments
i. Time priority is retained, but with an introduced delay in the flow
ii. All orders including non-colo orders will be impacted by the delay.
iii. Possibility of cascading effect of random delay on subsequent orders.

7.4 RANDOMIZATION OF ORDERS RECEIVED DURING A PERIOD (SAY 1-2 SECONDS)

i. Introduction

a. As per the mechanism, time-priority of the new / modified orders that would be received during
predefined time period (say 1-2 seconds period) is randomized and the revised queue with a
new time priority is then forwarded to the order matching engine.

b. Similar to the mechanism`s mentioned above, the said mechanism is expected to nullify the
latency advantage of the co-located players to a large extent that they get on the basis of physical
proximity to the trading platform and thereby, discourage latency sensitive active strategies.

ii. Benefits

a. Reduce Latency Advantage

The proposal is to reduce the latency advantage by changing the time priority of the new /
modified received in a certain time period. The revised queue with new time priority will be
forwarded to the order matching engine.

b. Profits available to fast traders will reduce

By making the timing of the market event uncertain, the profits available to fast traders will
also go down as they will not be guaranteed a “last-mover advantage”.

iii. Costs and Risks

a. Relevance of IOC orders

The mechanism defeats the time priority of the present system, and will vitiate the purpose of
IOC orders.

b. Increased uncertainty
The introduction of randomized order delay will in-crease the uncertainty and the adverse
selection costs of the trader, especially during volatile periods.

c. Lower liquidity
The new constraint could deter traders from trading. This will cause liquidity to migrate to
markets where such a feature does not exist, and will benefit the offshore exchanges at the cost
of domestic exchanges.

76
 d. Reduced cross-market arbitrage
The uncertainty regarding when an order will reach the exchange’s order matching could deter
the participants from executing cross-market arbitrage at short time frames. This can result in
in-creased disparity in prices across market segments.

e. Synchronizing randomization across all market venues and its global implementation
The hardest part to implement is the synchronized randomization across all market venues and
its global implementation. However, relaxing the synchronization requirement and the fully
global implementation would still likely yield significant benefits

f. Reduced market quality under the new mechanism

This may occur for example because of an increase in execution uncertainty and execution price
uncertainty for high frequency traders. However, low frequency participants are already subject
to these uncertainties and may in fact find it easier to monitor slower markets.

g. Reduce incentives to provide liquidity

Another negative implication is that by reducing the profitability of fast liquidity provision it
may reduce the positive externalities emerging from private incentives to provide liquidity.

iv. Evidence from Global Markets

a. ICAP’s EBS Market Matching Platform has introduced ‘Latency Floor’ that consists of a
random batching window of 1, 2 or 3 milliseconds, whereby all messages submitted within this
period are collected and then randomly released to the matching engine. The process is aimed
at ensuring that speed as a stand-alone strategy is not a pre-requisite for success on EBS Market.
b. Randomization is commonly used in opening and closing auctions especially in European
exchanges but there is no such study specifically of the effect of randomization

v. NSE’s view is that in absence of any substantiated advantage of randomization of already

randomly received orders and the uncertainty it brings in the order book the proposition does
not serve any objective.

vi. BSE’s view

Objective:
a. Collect orders over time intervals and taking up randomly for processing
b. Randomization logic is of utmost importance. Orders will be held into a common pool
and taken up for matching based on the randomized seed. This will be done on regular
intervals say 1 millisecond.

Comments
a. Akin to holding back orders for a duration, and randomly taking up for matching
b. Paradigm shift from an event based matching to timer based matching
c. Inherently has the ability to offer a chance to non-Colo orders even if the ratio of Colo
to non-Colo orders is huge.
d. Time priority is affected in this case.

77
7.5 MAXIMUM ORDER MESSAGE-TO-TRADE RATIO REQUIREMENT
i. Introduction

A maximum order-to-trade ratio requires a market participant to execute at least one trade for
a set number of order messages sent to a trading venue. The mechanism is expected to increase
the likelihood of a viewed quote being available to trade and reduce hyper-active order book
participation.

The mechanism is slightly different from ‘Order-to-Trade Penalty Rule’ implemented by the
stock exchanges in Indian securities market as the trader in the proposed case would not be able
to place such orders that further increase the ratio, after the limit is breached. As per the Order-
to-Trade penalty mechanism implemented by the stock exchanges in Indian securities market
penalty as per the prescribed slabs are imposed on the traders. There does not exist restrictions
on the placement of orders.

ii. Ambiguity

The proposal lacks clarity on the design. It is unclear at what time interval the ratio will be
calculated; whether it will be measured in real time at every point of time or whether it will be
accumulated over a certain time interval; whether it will be computed at the member level

iii. Benefits

a. Reducing the number of economically excessive messages

Receiving, handling and storing messages is costly for exchanges, brokers and regulators.
A ratio of orders-to-executions (OER) will reduce the number of economically excessive
messages. This, in turn, will reduce the need for exchanges, brokers and other market
participants to invest in costly capacity.

b. Increase Market depth

With fewer quote cancellations, the order book may be less active and traders may find it
easier to ascertain current prices and depths. An OER also may increase the likelihood of a
viewed quote being available to trade, partly because passive order submitters would focus
more on those limit orders with a higher probability of execution.

c. Curtail market manipulation strategies

Strategies such as such as quote stuffing, spoofing and layering are illegal but they are often
hard to detect. By limiting order cancellations, an order-to-execution ratio will reduce the
ability to implement these strategies.

iv. Costs and Risks

a. Reduced liquidity
If the requirement to maintain the maximum message to trade ratio becomes binding, the
imposition of this rule could constrain the traders from sending new orders or modify or
cancel existing orders. The inability to send new orders could reduce liquidity provision.
Such restrictions will inevitably result in stale quotes. This is likely to trigger a similar

78
 trigger of higher adverse selection costs, higher inventory risk for a liquidity provider, to
disincentive market participants to competitively place quotes at the best prices.

b. Withdrawals during volatile periods

The inability to send orders after breaching of the maximum limit will adversely affect the
ability of traders to provide liquidity during volatile periods.

c. Implementation costs: The implementation would require the exchanges to monitor the
message to trade ratio of each participant at a certain time interval. This will be an
additional cost for the exchange

d. Depth may be affected

If the order-to-trade ratio will be a non-binding constraint, the depth and bid-ask
competitiveness will not be affected. However, when a market participant does near the
maximum quote-to-trade ratio they will likely be more cautious about placing quotes given
they will be penalized if they withdraw too many quotes. This hesitance is most likely to
occur in volatile times, meaning that depth may be removed

e. Increase bid ask spread

Market participants will be more hesitant to place shares on the order book, they will also
be less likely to aggressively compete for posting quotes at the best bid and offer prices,
which will result in increased bid-ask spread

f. Less Transparent Limit Book

To find liquidity, traders often ‘ping’ or send small orders inside the spread to see if there
is hidden liquidity. Because such orders are typically cancelled, a binding OTR would
result in less pinging and, therefore, less information extraction at the touch. As a result,
more hidden orders will be posted, leading to a less transparent limit order book.

g. Where exactly to set any ratio and to what type of orders or traders it will apply
If the upper limit of the OER is small, then it will stifle legitimate activities and prevent
socially useful trading. For instance, ETFs and derivatives valuations may become
unaligned, leading to inefficient pricing. Because of this, the London Stock Exchange
(LSE) has an OTR of 500/1 for equities, ETFs and exchange traded products (ETPs), with
a high usage surcharge of five pence for equities and 1.25 pence for ETFs/ETPs.

If instead the upper limit is set high enough not to impinge on legitimate order strategies,
it may not have much impact on the market either (a point made by Farmer and Skourous
(EIA2:Minimum Resting Time and Transaction Order Ratio))

If the intent is to limit manipulative strategies, a specific charge for messages (and greater
surveillance) may be a better solution.

v. Evidence from Global Markets

a. There have been no published academic studies of OERs, and this greatly limits the ability
to gauge the costs and benefits of order activity restrictions in general and OERs in
particular. The commissioned study, (EIA18: Order to trade ratios and their impact on

79
 Italian stock market quality), investigates the effect of the introduction of an OER penalty
regime on the Milan Borsa on 2 April 2012. The authors’ preliminary findings are that
liquidity (spreads and depth) worsened as a result of this policy measure. They also find
that the effect is more pronounced in large stocks, although they acknowledge some issues
with their methodology.

b. There are a variety of actual market programs that provide some evidence of OER impact.
The ICAP has a monthly fill ratio (MFR) requiring that at least 10% of all quotes submitted
into the market must result in an execution.

c. LSE’s Millennium trading system has message throttling constraints and penalties for
excessive ordering strategies. Anecdotal evidence suggests that the LSE message policy
was not fully effective in that it gave rise to new patterns of trade in low-priced stocks. The
LSE has experimented with changes in pricing effective May 4, 2010 whereby, among
other measures, the threshold for the high usage surcharge for FTSE 350 securities
increased from an OTR of 100/1 to a ratio of 500/1 (which is still the figure in use at the
time of writing) along with high usage surcharge of five pence for equities and 1.25 pence
for ETFs/ETPs. The frequency of order book updates nearly doubled for a few months as
a result before coming down again.

vi. Evidence from Indian Markets

a. A similar instrument has been implemented twice before on orders in the Indian equity
derivatives markets:
 The first was by the NSE in 2009 (reduced in 2010) in order to manage excessive
placement of IOC orders causing bandwidth constraints at the exchange
 The second was by SEBI in 2012 (doubled in 2013) on orders that fell outside one
percentage band around the LTP.

b. Aggarwal et al. (2016) in their research paper “The causal impact of algorithm trading on
market quality ”analyse the impact of these interventions and find that such interventions
can be used effectively when:
 the objective is clearly stated and
 When it is effectively designed.

Research Paper:

Agarwal
Thomson_Causal Impact of algo trading on market quality.pdf

vii. NSE’s view is that this rule of order to trade ratio and penalties and severe action including not
able to participate in trading are already in place in India and the same is effectively working.

80
 viii. BSE’s view

Objective:-
a. Manage order to trade ratio in real time
b. Trading system will maintain the order and trade count and compute OTR in real time.
Based on the OTR rules, orders may be rejected\accepted.

Comments
a. Benefit in this case may be similar to that of minimum resting time
b. Present mechanism - disciplinary action for cooling off for first 15 minutes of trading on
the next trading day
c. Ensures that orders translate to trades more effectively

7.6 SEPARATE QUEUES FOR CO-LOCATION ORDERS AND NON-COLO ORDERS (2

QUEUES)

i. Introduction

a. With the view to ensure that stock brokers (and thereby the investors) who are not co-located
have fair and equitable access to the stock exchange’s trading systems, stock exchanges
facilitating co-location / proximity hosting shall implement an order handling architecture
comprising of two separate queues for co-located and non-colocated orders such that orders are
picked up from each queue alternatively. It is expected that such architecture will provide orders
generated from a non-colocated space a fair chance of execution and address concerns related
to being crowded-out by orders placed from colocation. The proposed architecture is as
described below:

b. Stock exchange shall identify and categorize orders as (a) orders emanating from servers of the
stock broker placed at the co-location / proximity hosting facility, and, (b) orders emanating
from other terminals / servers of the stock brokers.

c. Separate order-validation mechanism and a separate queue shall be maintained for each of the
aforementioned categories of orders.

d. A round-robin methodology shall be used to time-stamp and forward validated orders from the
two order-queues to the order-book, i.e., if an order is taken from the queue of orders emanating
from co-location / proximity hosting facility, then the next order shall be from the other queue.
In the event any of the order-queues are empty, orders can be sequentially taken from the other
queue till a valid order arrives in the empty queue.

e. As per the mechanism, separate queues and order-validation mechanism would be maintained
for co-lo orders and non-colo orders. Orders from queues will be taken up in the order-book in
round-robin fashion.

f. It may however be noted that the colocated participants would still be among the first to receive
the market data feeds due to their proximity to the trading platforms of the exchange and this
coupled with the capability to make trading decisions in fraction of seconds (by use of trading

81
 algorithms) would still provide the co-located participants the ability to quickly react to such
market data.

ii. Benefits
The benefits will be limited since the co-located participant would still be among the first to receive the
market data feeds due to his proximity to the exchange, and react to an opportunity because of
technological advantage. Thus, the benefits appear to be non-existent.

iii. Costs

a. Implementation costs
The implementation of the proposed mechanism would impose substantial costs on the market
and the economy.

b. Two markets for two sets of investors

The presence of two queues will result in two sets of prices at any given time for a security:
one coming from collocated participants and another for non-co-located participants. Since the
collocated participants will have the advantage of responding first to prices, these prices will
be the most efficient prices. This will lead to liquidity migrating to only the co-located market,
and killing off the market for non-co-located trading.

c. Withdrawal of liquidity
Such a mechanism has the potential to drive liquidity from domestic exchanges to offshore
exchanges where no such constraints would apply

iv. NSE’s view is that two queue proposition is effectively splitting the market into algo market vs
non algo market. Splitting the market may not be the regulatory objective. Exchanges may be
required to provide co-location service to all members transparently.

v. BSE’s Views

Objectives
a. Collect orders from Colo and non-Colo in separate queues for matching
b. Take up orders from each queue based on prescribed logic
c. Need to ensure that orders from non-Colo get a fair chance vis-à-vis orders from Colo
d. Apply exception conditions when order rates are skewed

Comments
a. Can potentially provide fair chance for non-Colo orders
b. Colo users will not have any visibility of the implementation
c. Algorithms can be applied to change the fairness when order rates between Colo and
non-Colo are skewed
d. No impact on trading members software

82
 7.7 REVIEW OF TICK-BY-TICK DATA FEED

i. Introduction

a. Tick-by-Tick (TBT) data feed provide details relating to orders (addition + modification +
cancellation) and trades on a real-time basis. TBT data feed facilitates a detailed view of the
order-book (such as depth at each price point, etc.).

b. At present, the exchanges provide TBT data feeds to any desirous market participant upon
payment of requisite fee.

c. Tick-by-Tick data feed is mainly subscribed by HFTs who coupled with their access to colocation
use such feeds to recreate the order-book and analyze the impact of execution.

d. TBT data feed is usually not availed by small players due to the feed being data-heavy (as it
includes details of all the order submissions, cancellations and modifications) and because of the
additional fee-component.

e. This has been viewed by a section of market participants to create disparity and inequality in
terms of access to data.

f. The proposal under examination is to provide ‘Structured Data’ containing Top 20 / Top 30 / Top
50 bids / asks, market depth, etc. to all the market participants at a prescribed time interval (or as
real-time feed).

g. The objective of the proposal is to adhere to the principle of market fairness by providing a level
playing field to the market participants irrespective of their technological or financial strength.

ii. Benefits

The information may be provided to all participants at the same time interval
However, a trader can benefit from real time feeds only if he has the infrastructure to assimilate the
huge data and act upon it. Inability to process this information will not result in any benefit from
elimination of TBT data and provision of real time feeds to all traders.

iii. Ambiguity

It is unclear as to how equality in access to data will help small traders, especially if these investors
are unable to take advantage of the tick by tick due to infrastructure requirements needed to process
large data sets.

iv. Costs

a. Can reduce the level of transparency if the data feed is anything other than real time feed
In this case, traders will not be able to see the price situation at every given point which will
adversely impact the price discovery process in the market.

83
v. Evidence from Indian Market

a. BSE uses multicast to disseminate TBT data(also known as EOBI) thereby ensuring
fairness for all consumers. Data is available free of cost to all colo members. Members need
to have suitable infrastructure to handle and process the TBT data. All colo members are
using this data. BSE can restrict dissemination of Full order book data immediately, if
mandated.

vi. In view of NSE, tick by tick data allows for more transparency to the market. Not giving or
restricting tick by tick data is restrictive and goes against the principles of transparency and will
affect market integrity and efficiency.

vii. BSE’s views

Comments
a) Allows members to access full order book.
b) Real time access to TBT data

84
Chapter-8

SURVEILLANCE SYSTEM
AT STOCK EXCHANGES

85
 SURVEILLANCE SYSTEM
Chapter 8
AT STOCK EXCHANGES

8.1 CURRENT SURVEILLANCE STRUCTURE AT NSE

Below given diagram illustrates the functional architecture

8.2 SURVEILLANCE FUNCTIONS COVERED AT NSE

These include online surveillance, surveillance actions and alert generation, as illustrated below:

86
8.3 CURRENT SURVEILLANCE MECHANISM ON NSE AT INTRA-DAY BASIS FOR ALGO
TRADES BASED ON ORDER TO TRADE RATIO
Introduction
• Applicable only for Algo Orders and trades and in FAO and CDS Seg.
• Order to Trade ratio = Order Messages / No. of Algo Trades
Note: - Order Messages = Orders Entered + Orders Modified + Orders Cancelled
Actions
• If ratio exceeds 500, member disabled for first 15 minutes on the next trading day. If ratio
exceeds 50, in more than 10 instances in previous 30 rolling days, member’s prop account
disabled for 1 hour on next trading day.
• Monetary penalty is being levied to members depending on their ratio and slabs.
Table 8.1: Monetary Penalty levied in last 3 Financial Years

Year Total Penalty Avg. Penalty Disablements

2014-15 6,96,097 58,008 9

2015-16 6,80,899 56,742 4

2016-17 1,76,963 58,988 Nil

Impact of current Surveillance mechanism on NSE at intra-day basis for algo trades
1. Impact on Order to Trade Ratio at Market level -

Table 8.2: Order to Trade Ratio at Market Level 2016

Month Apr-16 May-16 Jun-16 Jul-16 Aug-16 Sep-16 Oct-16 Nov-16

OTR 8.68 9.41 8.36 8.29 8.41 7.97 7.35 6.95

Note: Order to trade ratio has reduced from 9.41 to 6.95 levels
2. Impact on Member having highest Order to Trade Ratio

Month Apr-16 May-16 Jun-16 Jul-16 Aug-16 Sep-16 Oct-16 Nov-16

Ratio 214.72 213.28 280.53 266.13 222.78 222.58 182.59 151.79

Note: Order to trade ratio has reduced from 213 to 151 levels

87
3. Impact on Order to Trade Ratio in Northern Region–
OTR has improved from 32 to 25. Further letters were sent to 20 Members of which 14 Members
have improved their OTR. Exchange is pursuing the members who have not shown any
improvement.

After Meeting Before Meeting

25.44 32.10

8.4 RISK MANAGEMENT MECHANISMS FOR ALGORITHMS EMPLOYED BY THE

EXCHANGE

As prescribed by SEBI vide its circular CIR/MRD/DP/09/2012 dated March 30, 2012 trading members
desirous of placing orders using algorithms are required to ensure that the orders are routed through risk
management checks. For detailed risk controls, refer below attached Exchange circular no. 21793 dated
September 28, 2012.

Circular 21793 dated

September 28 2012.pdf

Exchange has various technical and functional checks to prevent rogue algos. There are order flow
controls or throttles for controlling the order flow from a member. Functionally, there are price bands
and price reasonability checks (dynamic price bands) in place to prevent cases of high volatility. Other
functional checks like Self Trade Prevention Check (STPC) and Reverse Trade Prevention Check
(RTPC) to prevent the same user(s) from manipulating trades, are also applied. Exchange also levies
‘Fair Usage Charges’ on members for inefficient execution of the order flow.

Refer Below for more details:

Annexure 5 - Sec 6
Q40 BSE Market Surveillance.pdf

88
Below is the summary of order-level and client -level algorithmic trading risk management:

S.N. Risk Control Applicability

At Individual Order Level:

Algo orders shall not be released in breach of the price bands
1 Price Check /dummy filters as defined by the Exchange in respective
segments.
Algo orders shall not be released in breach of order quantity
2 Quantity Check limit per order as defined by the Exchange in respective
segments

Algo orders shall not be released in breach of the “value per

3 Order Value Check order” ( combination of price and quantity checks) as defined
by the Exchange for the security in respective segments

Algo orders shall not be released in breach of the bad trade price
4 Trade Price Protection Check as defined by the Exchange for the security in respective
segments
Market orders emanating from Algo system shall not be
released beyond a pre-set percentage of LTP. The limit thus set
5 Market price protection
shall be less than the applicable circuit limits as prescribed
above
At Client Level :

Algo orders shall not be released in breach of Individual client

level cumulative open order value check at the client level.
Cumulative Open Order Value
6 Open order value for a client is the total value of its unexecuted
Check
orders released from trading members system. System shall not
permit the user to set “Unlimited values”

Algo to account for all executed, unexecuted, and unconfirmed

orders placed by the system before releasing further orders.
7 Automated Execution Check Algo system to have pre - defined parameters for an automatic
stoppage in event of Algo execution leading to a loop or a
runaway situation.

Algo orders shall not be released in breach of Net positions as

Net Position Vs. available
8 against the available margin defined by the trading member for
margin
the client.
RBI Violation checks for FII Algo orders shall not be released for security restricted by RBI
9
Restricted stocks from time to time
Algo orders shall not be released in breach of market wide
10 MWPL violation check position limit as defined by the Exchange for the security in
respective segments
Algo orders shall not be released in breach of position limit as
11 Position Limit Checks
defined by the trading member for the client

89
 Algo orders shall not be released in breach of overall trading
12 Trading Limit Checks
limits as defined by the trading member for the client

Exposure Limit check at

Algo orders shall not be released in breach of exposure limit as
13 individual client level and at
defined by the trading member for the client
overall level for all clients

8.5 SEPARATE RISK MANAGEMENT MECHANISMS FOR HFT

Trading members are not required to share the business logic of algo as it is a proprietary information.
Further, SEBI has not prescribed any specific requirements for HFT as it is a part of Algos. However
members may build risk management systems over and above the ones prescribed by SEBI

Risk management controls are implemented for all order flows. Users of HFT have the HF ids, while
the non algo users are provided the LF ids. The order flow rate for HF ids is controlled at millisecond
level while the LF ids rate is controlled at seconds level

8.6 SURVEILLANCE MECHANISMS IN PLACE TO CATCH HARMFUL HFT ACTIVITIES

The Exchange amongst other alerts, generates and analyses following type of alerts specifically from
algo trading behaviour:

 Order Spoofing,
 Quote Stuffing,
 Marking the close,
 Large Cancelled orders
 Large unexecuted orders

Surveillance processes are applied to all trades irrespective of the origin or mode in various jurisdictions
including in India. At BSE, HFTs are generally routed through Co-Location which has a unique
identifier. The Exchange Surveillance system based on the pre-determined criteria generates alerts and
reports for monitoring the trading for Algo as well as Non-Algo orders. The Algo orders carry separate
indicator as an identifier.

Exchange has various technical and functional checks to prevent rogue algos. There are order flow
controls or throttles for controlling the order flow from a member. Functionally, there are price bands
and price reasonability checks (dynamic price bands) in place to prevent cases of high volatility. Other
functional checks like Self Trade Prevention Check (STPC) and Reverse Trade Prevention
Check(RTPC) to prevent the same user(s) from manipulating trades, are also applied. Exchange also
levies ‘Fair Usage Charges’ on members for inefficient execution of the order flow.

90
8.7 DO SURVEILLANCE MECHANISMS CHANGE IN PERIODS OF HIGH VOLATILITY

Prior to any major event Surveillance calls for a meeting of members wherein members share their
views, trends prevailing in the market and their risk containment measure. The same process is carried
out along with SEBI.
Further Exchange has put in place various systems/ measures such as price bands, trade execution range,
dynamic price bands, risk reduction mode etc. which helps mitigate risk and helps in orderly functioning
of the market.

High volatility in the market can be witnessed on account of specific micro & macro events and may
lead to increase in inflow of orders/trades and price volatility in the contacts/securities being traded at
the Exchange.

The Surveillance systems at BSE is regarded as mission critical system and the same has been tested at
their peak capacity and are capable of handling instances of high volatility.

Further, to address in the event of high volatility, the Exchanges in India have put in place a mechanism
of Market Wide Circuit Breaker as per the Regulatory mandate to be applied across the Exchanges in
coordinated manner. These circuit breakers when triggered bring about a coordinated trading halt in all
equity and equity derivative markets nationwide i.e. across the Exchanges.

8.8 KILL SWITCHES

Kill switch facility allows the member to cancel all orders sent to and pending at the Exchange with a
single function. All orders entered by the specific user invoking the said facility and if pending at the
Exchange, shall be cancelled.

NSE has deployed kill switch functionality in trading vide its circular no. 26337 dated March 31,
2014.(refer below)

Circular 26337 dated

March 31 2014.pdf

Members are required to terminate the dysfunctional algorithm at their end using kill switch. This can
be either manual or automated as developed by the member.

Till date, the kill switch is not deployed by the Exchange. If Exchange detects rogue behavior of an
algo, Exchange would inform the member to deploy the kill switch.

91
8.9 NEED BY EXCHANGES TO ENHANCE ITS SURVEILLANCE SOFTWARE/ PROGRAMS
It can be concluded that NSE and BSE has robust broader level Risk Management and Surveillance
systems in place that are capable of identifying abnormal market activity and trade manipulation for
necessary actions.

However, the exchanges need to control and curb HFT malpractices through the introduction of
advanced Surveillance mechanisms and technology, some of which are discussed below:

 SEBI should have a unified framework for exchanges to manage surveillance on the algorithmic
trading front.
 Currently each exchange has its own methods and levels of sophistication to manage
surveillance. Harmonization of surveillance mechanism would bring about uniformity in
exchange action towards harmful HFT.
 There is a definite need to invest in advanced technology to automatically detect harmful HFT
and market manipulative trends/algorithms.
 Currently, exchanges hardly have advanced mechanisms to detect harmful HFT – there is a
need for technological advancement wherein a team of advanced algo specialists be setup and
appropriate infrastructure be invested in, which build detection algorithms to catch consistent
market manipulative behaviors of harmful HFT algorithms and also consistently mine market
manipulative trends emerging from order-to-trade ratios.
 Currently, there is no mechanism to check the algo submitted by any trading member. From
surveillance, if consistently high order-to-trade ratios are detected for certain trading members,
there should be norms to check the algorithm logic of the member. There needs to be regulatory
framework allowing this as well as enough algo experts on the exchange front who understand
algorithms.

92
 Chapter-9

ALGO TRADING AND EXCHANGE

APPROVALS

93
 ALGO TRADING AND
Chapter 9
EXCHANGE APPROVALS

9.1 PROCEDURE FOR GAINING ALGORITHMIC TRADING APPROVAL

A) Following are the pre-requisite for trading members seeking algo trading approval:

o Submission of Undertaking
o Meeting Base Minimum Capital (BMC) requirement of Rs.50 lakhs
o Testing of Software in UAT, Test and Mock environment
o Auditor certificate

B) Below is the approval process:

o Submission of application: Trading members desirous of seeking approval for algo

trading are required to make an application to the Exchange electronically.
o Demonstration of Software: Trading members shall be required to give demonstration
of the Automated Risk Management features as mandated by SEBI /Exchange.
o Approval: On verification of the documents and fulfilment of the conditions as
satisfactory and meeting SEBI / Exchange minimum requirements, the Exchange
grants permission to the trading member.

As per SEBI circular CIR/MRD/DP/ 09 /2012 dated March 30, 2012 Exchange has designed a
comprehensive application form and undertaking to be provided by members desirous of seeking
approval of algos from Exchange. Trading Members desirous of using algo trading facility for Prop /
Client trades are required to seek approval from the Exchange. Members desirous of getting their algos
approved should make an application with details of the algo and version. They also need to either
participate in the mock, or get the algos verified in the Exchange simulation environment. They also
need to get their algos certified by a CISA/DISA certified auditor of having tested the same with respect
to the Exchange risk management requirements. Subsequently, the certified algo is placed before the
exchange Algo committee. Based on the approvals from the exchange Algo committee, the algo is
allowed for use by the member.

9.2 PROCESS FOR FIRST TIME ALGO APPROVAL OR ADDITIONAL SOFTWARE FROM
DIFFERENT IT VENDOR:

1. Trading member submits the following documents as per checklist :

i. Algo application form

ii. Auditor Testing Report

iii. Auditor Covering letter and system audit report

94
 iv. Software & Algo Undertaking (one time submission. Not required if given
earlier)*

2. On the receipt of complete documents Exchange verifies that whether the trading member has
ensured that the such algo facility adheres to following automated risk management checks

i. Fat Finger' check - Any single order does not carry abnormally large quantity
beyond pre-defined limit

ii. Any single order does not carry abnormally large value beyond pre-defined limit

iii. The price of any single order should not be more than the applicable price bands
including dummy price bands defined by the Exchange for security/contract.

iv. All orders generated by Algorithmic trading products are tagged with appropriate
location code(s) as specified by the Exchange and such location code(s) are
registered with the Exchange

v. The stock broker has maintain logs of all trading activities to facilitate audit trail.
The stock broker has maintain record of control parameters, orders, trades and
data points emanating from trades executed through algorithm trading.

vi. All orders generated by Algorithmic trading products are adhering to limit of
maximum number of permissible orders per second as may be specified by the
broker at his end.

vii. Before pumping in further orders, the algo checks if the orders already released
by it (executed, unexecuted, unconfirmed) are sufficient with respect to strategy
of the algo.

viii. The system has sufficient security features including password protection for the
user ID, automatic expiry of passwords as the end of a reasonable duration and
re-initialization of access on entering fresh passwords. This functionality is
verified that the system has enough security features to ensure that the access to
the software is only available to authorized persons at the members end.

ix. Dysfunctional algos are broadly defined as algo that does not work as per defined
logic or does not conform to the risk management checks defined by the broker or
affects orderly trading and market integrity on account of its potential misfiring.

1 - Annexure 1a Algo Trading

(Sec1 Q 3 Algo approval process).docx
Application Form.doc

95
9.3 RULE FOR CHECKING ALGOS AT A LOGIC LEVEL BY THE EXCHANGE.
Is there any kind of versioning system to capture the algorithms used by participants in a secure
location?

1. Trading members are not required to share the business logic of algo as it is a proprietary
information.
2. Versions as reported by trading members during the approval process are captured by the
Exchange

Exchange does not examine the business logic of Algorithms. However each algo version is certified.
The onus is on the Trading Member to report any subsequent change/s in the version of ALGOs being
used. Any major change/s in the business logic of the ALGO being used or a new ALGO being proposed
for use shall require the Trading Member to seek fresh approval from the Exchange before use.

9.4 MECHANISM TO CAPTURE CHANGE IN ALGOS AT PARTICIPANTS END BY THE

EXCHANGE

A) As prescribed by SEBI vide their circulars CIR/MRD/DP/09/2012 dated March 30, 2012 and
CIR/MRD/DP/24/2013 dated August 19, 2013, trading members are required to seek approval
before effecting material changes to the algo trading software already approved by the
Exchange. The process for approval of modifications to Algos remains same as for new Algo
approval explained above in response to point 3.

Further, as prescribed by SEBI vide its circular CIR/MRD/06/2014 dated February 07, 2014
for approval of change/modification in the trading software arising out of the bug in software,
change undertaken in software/system pursuant to change prescribed by exchange or software
purchased from a software vendor that has already been tested in mock environment, Exchange
adopts the following algo approval process:

i. Submission of application: Trading members desirous of seeking approval for

modification in algo are required to make an application to the Exchange electronically
along with auditor certificate confirm testing of software in UAT, test and simulated
environment of Exchange.

ii. Demonstration of Software: Trading members shall be required to give demonstration of

the Automated Risk Management features as mandated by SEBI /Exchange.

iii. Approval: On verification of the documents and fulfilment of the conditions as

satisfactory and meeting SEBI / Exchange minimum requirements, the Exchange grants
permission to the trading member.

B) As prescribed by SEBI vide its circular CIR/MRD/DP/16/2013 dated May 21, 2013 and

96
 CIR/MRD/DMS/34/2013 dated November 19, 2013 as part of periodic system audit required
to be conducted by an independent system auditor, SEBI prescribed terms of reference (TOR)
cover the need to certify that change process has been followed. For details of terms of reference
refer sub point 17 of attached file “TOR.doc”

As a part of ALGO approval process, trading member has to submit an ALGO Undertaking wherein
the Trading member undertakes to unconditionally and irrevocably comply with various terms and
conditions pertaining to the use of Algorithmic Trading facility.

Additionally, half yearly system audit has been prescribed for stock brokers who use Algorithmic
Trading or provide their clients with the facility of Algorithmic Trading as per SEBI Circular
CIR/MRD/16/2013 dated May 21, 2013. Further SEBI vide its circular CIR/MRD/DMS/ 34 /2013 dated
November 6, 2013 has prescribed guidelines for Stock Broker Audit Framework.

The auditor in its report shall specify compliance / non-compliance status with regard to areas
mentioned in ToR (terms of reference) mentioned in the SEBI circular. The auditor shall also take into
consideration the observations / issues mentioned in the previous audit reports and cover open items in
the report. This shall ensure that any changes to ALGO trading facility are captured and non-
conformities are reported.

Algo SEBI_System audit Sebi circular on Algo

Undertaking.doc Framework_NOV6, 2013.pdf
System audit_MAY21, 2013.pdf

9.5 BEST PRACTICES INTERNATIONALLY FOR ANY MEMBER WANTING ACCESS TO

ALGORITHMIC TRAINING

As per the understanding gathered from market participants who operate in multiple jurisdictions,
information regarding process/practices followed by international exchanges is not available in public
domain. However, as given to understand by the trading members, in most Asian exchanges there are
no separate process/requirements for algo trading. In some of the exchanges, for approval of member’s
system, members share the testing results and logs generated from the exchange provided test cases.
Test cases cover the overall risk management prescribed by the exchange. Exchange verifies the testing
results and upon satisfaction provides necessary approval.

Based on feedback from some vendors who are providing products to international exchanges, there is
no algo certification required in other exchanges. However there is vendor certification required. Eg.
Vendors are provided a facility to test their products in CME, based on which vendors are approved.

97
9.6 QUALIFICATION REQUIREMENT FOR MEMBERS INTENDING TO DO ALGO TRADING

As prescribed by SEBI vide circular CIR/MRD/DRMNP/36/2012 dated December 19, 2012, any
trading member intending to use algos is required to furnish the highest applicable Base Minimum
Capital of Rs. 50 lakhs on which the member is not provided any trading exposure. At the time of
registration as a member itself, Trading members are required to fulfil certain criteria prescribed by
SEBI/ Exchange such as minimum capital, minimum networth, minimum security deposits etc. Further,
designated directors of the member are required to fulfill certain criteria like minimum age, education
qualification, total experience in dealing in securities etc. Further, the member is required to appoint an
appropriately qualified compliance officer holding a SEBI mandated certificate. Also, for issuance of
user id for trading purpose, a valid certificate issued through SEBI approved certification programme
is required to be furnished by trading member.

Currently members intending to do algo trading require BMC (Base Minimum Capital) requirement of
Rs.50 lakhs (Financial requirement) which is required to be maintained by member with the Exchange
(as per SEBI requirement). This is higher than the normal BMC requirement for Trading members not
using ALGO trading facility.

9.7 EXCHANGE ENSURE ROBUSTNESS OF ALGORITHMS ON AN ONGOING BASIS

As per SEBI circular CIR/MRD/DP/24/2013 dated August 19, 2013 algo members are required to test
their algorithms in mandatory monthly mock sessions scheduled by Exchange. Exchange checks
member's participation subsequently and any non-compliance is considered for appropriate action.
Additionally, as per SEBI circular CIR/MRD/DP/16/2013 dated May 21, 2013 and
CIR/MRD/DMS/34/2013 dated November 06, 2013 algo members are required to conduct half yearly
system audit as per SEBI prescribed Terms of Reference (TOR).

TOR.docx

Are results of these simulations validated to identify robustness of existing algorithms? As per
SEBI, adherence to monthly mock participation is part of half yearly system audit requirement. For
details, refer sub point 17 of attached file “TOR.doc”

Algo approvals are granted to Trading Members based on the certification provided by qualified system
auditors. The auditors are required to certify the Algos based on the checklist provided by the Exchange
(as per SEBI guidelines). In addition, the certification by the qualified system auditors, also confirms

98
the algos compliance with Exchange risk management systems during mock trading sessions or testing
in simulation environment.

Exchange has various technical and functional checks to prevent rogue algos. There are order flow
controls or throttles for controlling the order flow from a member. Functionally, there are price bands
and price reasonability checks (dynamic price bands) in place to prevent cases of high volatility. Other
functional checks like Self Trade Prevention Check (STPC) and Reverse Trade Prevention Check
(RTPC) to prevent the same user(s) from manipulating trades, are also applied. Exchange also levies
‘Fair Usage Charges’ on members for inefficient execution of the order flow.

Exchange does a daily check on Order to Trade Ratio (OTR) for Equity Derivative/Currency Derivative
and levies charges for High Order to Trade ratio (OTR) in Algorithmic Trading for Equity Derivative
segment (circular no : 20130529-6) and Currency Derivative segments (circular no : 20141010-26).

Exchange does monthly checks for the following:

i. Fair Usage Charges levied on members BSE Equity, Equity Derivatives and Currency Derivatives
circular no: 20161027-14.
ii. Order Flow Rate in Equity Segment circular no*: 20161019-35, Currency derivative Segment
circular no*: 20160930-3, Equity Derivative Segment circular no*: 20160503-22.
*Circulars are attached in point ii

Are results of these simulations validated to identify robustness of existing algorithms?

All checks done in mock or simulation environment are validated by system auditors and certified,
before making an application with the Exchange.

9.8 ALGO TESTING LAB

Algo Lab is an optional facility developed and provided by the Exchange. It provides a near live market
simulator designed for testing of algorithms. It provides actual market feed for testing and acts as a
comprehensive analytics environment for strategy refinement. Algo Lab environment provides a replica
of actual market conditions by creating a similar co-located environment for dealers to test their
strategies/algorithms against a dedicated market feed. Some of the benefits of algo lab is given below:-

 Assess the performance of algorithms

 Test new strategies without incurring financial risk
 Train new dealers

The algo test lab is a test bed for trading members to test their algos before moving to the Exchange
Simulation/Mock/ live environments. This is helpful to members to validate the logic of algos in
simulated market conditions, based on BSE market data.

99
Chapter-10

CONCLUSION

100
 Chapter 10
CONCLUSION

10.1 ALGO TRADING AND ITS COMPOSITION IN INDIAN MARKETS

a. Algorithms have become such a common feature in the trading landscape that it is unthinkable
for a broker not to offer them because that is what clients demand. High-frequency trading
(HFT) is a subset of automated trading. Here, opportunities are sought and taken advantage
of on very small timescales from nanoseconds up to milliseconds.
b. Composition
We found that around 50% plus of total orders at both NSE and BSE are algo trades -client
side. Prop side algo trades are 40% plus of total orders placed at both the exchanges. 80%
plus orders are generated from colo at BSE ( of total algo orders)

10.2 EXTENT OF ALGO TRADING IN INDIAN & FOREIGN MARKET AND ORDER TO
TRADE RATIO

a. In India, Algorithmic trading to Exchange turnover has stabilised around 47%. In developed
markets it stands at about 80%.
b. Order to Trade Ratio in Indian Market
In the year 2016-17, order to trade ratio for NSE across all segments was 11.2. It has increased
from 7.07 in 2014-15. In BSE: Order to trade ratio for equity segment has decreased to 218
(2016-17) from 321.57(2014-15), for equity derivatives decreased to 3597 (2016-17) from
8162 (2014-15), for interest rate derivatives decreased from 884 (2014-15) to 132(2016-17)
however for currency derivative it has increased from 111(2014-15) to 195(2016-17). In case
of high order to trade ratio, NSE makes calls and alert trading members. BSE has issued
circulars to keep a check on high order to trade ratios. Penalty is imposed on both the exchange
for high to trade ratio.

10.3 ADVANTAGES DISADVANTAGES OF ALGO TRADING, COLOCATION AND HFT

Algo trading, colocation and HFT offer various advantages and disadvantages. It is observed that with
algo trading and HFT there have been improvements in transactions costs, volatility, and buy-sell
imbalance. Market prices have become more efficient and they have facilitated price discovery.
Colocation also reduced latency and levels the playing field among competing HFT market makers.
However, Technical sufficiency and resources are required for advanced technology. Lack of control
has led to systemic risks. Fat finger or faulty algorithms can cause huge deviations from healthy prices
HFT is playing against investors who are willing to invest fundamentally. HFT firms leverage special
services such as co-location facilities and raw data feeds. The odds of a sudden liquidity drain go up.
HFT can give rise to price fluctuations and short term volatility. Colocation facility can be expensive
and gives rise to market inequity

10.4 MARKET MANIPULATION USING ALGO TRADING

It has been proved in the past that Algo tading and HFT can used to manipulate markets using techniques
like quote stuffing, layering (spoofing) and momentum ignition. Evidence suggests that because of
quote stuffing stocks experience decreased liquidity, higher trading costs, and increased short term
volatility. Layering can impact performance and fill rates. Momentum ignition events can result in
massive price moves backed by false volume

101
10.5 MEASURES ADOPTED BY SECURITIES MARKET REGULATORS IN DIFFERENT
COUNTRIES

Minimum resting time, frequent batch auctions, random speed bumps or delays, randomization of orders
during a period (1-2 seconds), max order message to trade ratio requirement, market maker pricing are
some of the measured adopted. HFT transaction tax is implemented in France and Italy. The SEC has
undertaken following steps: Market Access Rule, Regulation SCI and registration of entities. In May
2013, Germany enacted the Act on the Prevention of Risks and Abuse in High-Frequency Trading (HFT
Act)

10.6 SURVEILLANCE METHODS AT NSE & BSE

Currently both NSE and BSE have its own methods and levels of sophistication to manage surveillance.
However, in our view harmonization of surveillance mechanism would bring about uniformity in
exchange action towards harmful HFT There is a definite need to invest in advanced technology to
automatically detect harmful HFT and market manipulative trends/algorithms. Exchanges hardly have
advanced mechanisms to detect harmful HFT.

10.7 PROS AND CONS OF REGULATORY MECHANISMS BEING CONSIDERED BY SEBI:

a. Minimum resting time for orders

Pros

 It offers advantages like :

 More stability in limit order book
 Reduces fleeting order
 Can increase the likelihood of a viewed quote being available to trade
 Reduces the excessive level of message traffic currently found in electronic markets

Cons

 It may lead to more order being pumped into the system and longer queues and waiting time
 It may raise transaction costs as well as volatility

b. Frequent batch auctions

Pros

 It will lead to reduction of the speed of trading and the elimination of the arms race for speed
 Eliminate Sniping

Cons

 It may impede price discovery and increase execution risk

 It may reduce liquidity and incentives to supply liquidity

Implementation in Markets

 Taiwan Stock Exchange (TWSE) used to have continuous auction mechanism

102
c. Random speed bumps or delays in order processing/ matching

Pros

 All orders including non-colo orders will be impacted by the delay

 nullify latency advantage and
 stops arms race for speed

Cons

 However a cushion can be provided against bumps

 It can cause reduced / withdrawal of liquidity

Implementation in Markets

 TSXA – Toronto Stock Exchange (1-3 ms) and ParFX – interdealer OTC broker (20-80ms)
impose randomized order processing

d. Randomization of orders received during a period (1-2seconds)

Pros

 It inherently has the ability to offer a chance to non-Colo orders even if the ratio of Colo to non-
Colo orders is huge. It reduce Latency Advantage, profits available to fast traders will reduce

Cons

 It will lead to increased uncertainty and lower liquidity

 It reduced cross market arbitrage

Implementation in Markets

• ICAP EB (wholesale FX electronic trading platform) Market Matching platform has introduced
Latency floor

e. Maximum order to trade ratio requirement

Pros

 This will ensure that orders translate to trades more effectively

 It Reduces the number of economically excessive messages
 Increase Market depth
 Curtail market manipulation strategies

Cons

 It may lead to reduced liquidity

 Withdrawals during volatile periods
 Increase bid ask spread

103
 Implementation in Markets

 NSE and BSE are currently implementing maximum order to trade ratio and penalties
 The ICAP has a monthly fill ratio (MFR)

f. Separate queues for co-location and non-co-location orders

Pros

 This can potentially provide fair chance for non-Colocation orders

 No impact on trading members software
 Algorithms can be applied to change the fairness when order rates between Colo and non-Co-
location are skewed

Cons

 This will lead to two markets for two sets of investors

 Withdrawal of liquidity

g. Review of tick by tick data feed

Pros

 More transparency
 Members to access full order book and
 Real time access to TBT data

Cons

 It can reduce the level of transparency if the data feed is anything other than real time feed

104
REFERENCES

Chapter-10

105
 REFERENCES

Regulatory studies/Discussion Papers

1. Govt of science paper” Future of Computer Trading in Financial Markets”:

https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/289040/12-1061-
dr28-high-frequency-trading-impact-on-market-efficiency.pdf
2. SEBI’s Discussion Paper Strengthening of the Regulatory framework for Algorithmic Trading
&Co-location http://www.sebi.gov.in/cms/sebi_data/attachdocs/1470393485587.pdf
3. https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/289047/12-1064-
eia2-minimum-resting-times-and-transaction-order-ratios.pdf
4. https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/289041/12-1065-
eia3-minimum-resting-times-vs-call-markets.pdf
5. http://people.stern.nyu.edu/jhasbrou/SternMicroMtg/SternMicroMtg2016/Papers/36.pdf
6. https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/289046/12-1077-
eia18-order-to-trade-ratios-impact-italian-stock-market.pdf
7. https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/289047/12-1064-
eia2-minimum-resting-times-and-transaction-order-ratios.pdf
8. http://asic.gov.au/regulatory-resources/markets/market-structure/dark-liquidity-and-high-
frequency-trading/.
9. http://www.iosco.org/library/pubdocs/pdf/IOSCOPD354.pdf

Research Papers

1. https://academic.oup.com/qje/article/130/4/1547/1916146/The-High-Frequency-Trading-Arms-
Race-Frequent
2. http://people.stern.nyu.edu/jhasbrou/SternMicroMtg/SternMicroMtg2016/Papers/36.pdf
3. https://www.researchgate.net/profile/HsiuChuan_Lee/publication/256018668_Auction_Designsan
d_Futures_Price_Behavior_Evidence_from_the_Taiwan_Futures_Market/links/55ece78d08aeb65
16268cdbd.pdf
4. http://mitsloan.mit.edu/groups/template/pdf/Zhang.pdf
5. http://www.ecmcrc.org/Articles/The_Impact_of_Co-
Location_of_Securities_Exchanges%E2%80%99_and_Traders%E2%80%99Computer_Servers_o
n_Market_Liquidity.pdf
6. http://people.stern.nyu.edu/jhasbrou/SternMicroMtg/SternMicroMtg2014/Papers/TongHFT20131
117.pdf
7. http://www.people.hbs.edu/kfroot/oldwebsite/cvpaperlinks/herd_on_the_street.pdf
8. https://www.business.msstate.edu/magnolia/pdf/quote-stuffing.pdf
9. http://febs2016malaga.com/wp-content/uploads/2016/06/7AnEmpirical.pdf
10. https://www.capgemini.com/resource-file-
access/resource/pdf/High_Frequency_Trading__Evolution_and_the_Future.pdf
11. https://www.edwards.usask.ca/csfm/_files/papers2011/4c-
Subsidizing%20Liquidity%20The%20impact%20of%20make-
take%20fees%20on%20market%20quality,%20K.%20Malinova,%20A.%20Park.pdf
12. https://ir.canterbury.ac.nz/bitstream/handle/10092/5693/12633224_Boyle.pdf?sequence=1
13. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1722924
14. https://papers.ssrn.com/sol3/papers2.cfm?abstract_id=1752843
15. http://people.stern.nyu.edu/jhasbrou/SternMicroMtg/SternMicroMtg2016/Papers/36.pdf
16. http://www.igidr.ac.in/pdf/publication/WP-2014-023.pdf
17. https://ifrogs.org/PDF/201608_responseFRG.pdf

106
 QUESTIONNAIRE
E
Chapter-10

QUESTIONNAIRE ON ALGORITHMIC TRADING AND

CO-LOCATION SERVICES

107
 QUESTIONNAIRE ON ALGORITHMIC TRADING AND
CO-LOCATION SERVICES

Section 1: Algorithmic Trading

1. What is the composition of algo trading/ high frequency trading taking place in your exchange
– client and proprietary?
3. How many of the algo orders come from co-location?
4. What is the procedure for gaining algorithmic trading approval?
5. Are algorithms checked at a business logic level by the exchange? Is there any kind of
versioning system to capture the algorithms used by participants in a secure location?
6. Is there any mechanism to capture change in algorithms at the participants end? How do you
detect if the participant is changing an algorithm?
7. Is it the responsibility of the participant or the exchange pro-actively checks versioning of
algorithms used by participants?
8. Internationally, what are best practices for any member wanting access to algorithmic training?
What are the qualification requirements?
9. Do we have any such qualification requirement for members intending to do algo trading? How
can we improve training and qualification of members?
10. How does exchange ensure robustness of algorithms on an ongoing basis? For instance, are any
simulated/mock trading sessions done?
11. Are the results of these simulations validated to identify robustness of existing algorithms or
only for macro risk management checks?
12. How does an algo testing lab help in active risk management by the exchange? If not, what is
the purpose of the algo lab?

Section 2: Algo market demographics and order-to-trade ratio

13. How many algorithmic trading players are currently registered? What is the ratio of active
participants?
14. Who are the top 20 algo participants and what is the contribution to average daily turnover?
15. What is the average order to trade ratio for all algo participants in the last 3 years (year-wise
bifurcation)?
16. What are the ratios of order-to-trade for the top 10 participants (by turnover) across the last 3
years (year-wise bifurcation)?
17. What are the order-to-trade ratios for efficient vs inefficient members across the last 3 years
(year-wise bifurcation)?
18. What has been the trend of order-to-trade ratio over the last 3 years?
19. Has the exchange been pro-active in controlling the order-to-trade ratio? If yes, how?
20. Are there any penalties or structures in place to curb high order-to-trade ratios?
21. How do you penalize repeat offenders of high order-to-trade ratio?
22. What causes a very high order-to-trade ratio?
23. Specifically, what category of HFT algorithms could lead to high order-to-trade ratios?
1. Section 3: Co-location
24. What are the costs of setting up co-location servers?
25. What are the additional costs that would be typically incurred by members for setting up co-lo
(apart from rack space)?

108
 26. What are the requirements to be met by members for algorithmic trading? And co-location?
27. Are there different types of co-location services for different types of participants?
28. If yes, what are the differentiated costs for different services?
29. Are there any benefits to members to setup co-location services at multiple exchanges? If yes,
explain how?

Section 4: Technology for co-location

30. Is there any technological standardization for co-location services? Or are there differentiated
levels (eg, rack space, wire connecting co-lo servers to exchange, etc)?
31. If yes, explain the differentiated levels of service along with associated costs.
32. What is the length of the wire used? What is the quality of wire used? Is it same for all members?
33. Which system (unicast/multicast) has been in existence for the last 5 years? Kindly explain the
workings of the system (eg TCPIP sequential access for unicast).
34. Can some participants gain undue advantage due to sequential access due to early log-in?
Explain using examples.
35. What is the difference between main server and back-up server? Is the back-up server used as
a parallel processor sometimes? Is there any preferential treatment on back-up server and
additional charges if applicable?
36. If yes, why access given to only a few participants on back-up server and not to all participants
in co-lo? If given, why?

Section 5: How do you capture data (order/trade information)

37. What is the format for capturing raw data (order/trade information)? Kindly share the raw data
format with us (converting into understandable format giving detailed description of each of
the fields)
38. Is this a standardized format that is followed internationally? Please describe the format. Or is
there a standardized unified format across all exchanges in India?
39. What is the international best practices for unified data formats? How will this help in Indian
markets? What are the benefits of data format harmonization?
1. Section 6: Risk Management and Surveillance
40. What are the best practices internationally for risk management in Algorithmic trading?
41. What are the risk management mechanisms for algorithms employed by the exchange?
42. Are there separate risk management mechanisms for HFT?
43. What are the surveillance mechanisms in place to catch harmful HFT activities? List the
mechanisms as well as explain them.
44. How do the surveillance mechanisms change in periods of high volatility?
45. What are kill switches? Who has control of them – exchange or members? Is it manual or
automated?
46. Has kill switch been ever employed in trading?

Section 7: Curbing harmful HFT algorithms

47. Kindly share your views on SEBI’s discussion paper. Which of the discussed points are relevant
for Indian markets?
48. Also, are there additional measures (not discussed in the paper) that could be introduced to curb
harmful impact of HFT?

109
49. If yes, what are the historical precedents internationally for the same?
50. According to your knowledge, what are the harmful HFT algorithms that might cause market
disruption such as high intra-day volatility and high order-to-trade ratios?
51. On the above question, how can you think these harmful HFT activities can be curbed? What
do you think are some international best practices that can be employed?
52. How do you think having separate queues, one for algo trades and the other for non-algo trades,
help in curbing harmful HFT?
53. How as an exchange would you identify such harmful HFT practices and algorithms?

110