DATA ANALYTICS IN
FOOTBALL
Data Analytics in Football provides students, researchers and coaches with a firm
grounding in the principles of modern performance analysis. It offers an insight
into the use of positional data, exploring how they can be collected, modelled,
analysed and interpreted. Introducing cutting-edge methods, the book challenges
long-held assumptions and encourages a new way of thinking about football
analysis.
Based on data collected from the German Bundesliga and the UEFA Champions
League, the book seeks to define the role of positional data in football match
analysis by exploring topics such as:
• What is positional data analysis and how did it emerge from conventional
match analysis?
• How can positional data be collected and which technologies can be used?
• What are the benefits of a data-driven approach to decision making in football?
• What Key Performance Indicators based on positional data should be used?
• How can traditional match analysis be complemented by using positional data
and advanced KPIs?
• How can these new methods evolve in the future?
Accessibly written, packed full of examples from elite football and supplemented
with expert interviews (Ralf Rangnick, Urs Siegenthaler and others), Data Analytics
in Football is a thought-provoking, rigorously evidence-based guide to the use of
data analytics in football performance analysis. As such, it is a vital resource for any
student, researcher or coach interested in performance analysis and skill acquisition,
or anyone interested in football more generally.
Daniel Memmert is Professor and Executive Head of the Institute of Training
and Computer Science in Sport at the German Sport University Cologne,
Germany.
Dominik Raabe is Research Assistant at the Institute of Training and Computer
Science in Sport at the German Sport University Cologne, Germany. He is also
studying Scientific Computing (MSc) at the Institute of Mathematics at the
Technical University Berlin.
DATA ANALYTICS
IN FOOTBALL
Positional Data Collection,
Modelling and Analysis
Daniel Memmert and Dominik Raabe
First published in English 2018
by Routledge
2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN
and by Routledge
711 Third Avenue, New York, NY 10017
Routledge is an imprint of the Taylor & Francis Group, an informa business
© 2018 Daniel Memmert and Dominik Raabe
The right of Daniel Memmert and Dominik Raabe to be identified as authors
of this work has been asserted by them in accordance with sections 77 and 78
of the Copyright, Designs and Patents Act 1988.
All rights reserved. No part of this book may be reprinted or
reproduced or utilised in any form or by any electronic, mechanical,
or other means, now known or hereafter invented, including photocopying
and recording, or in any information storage or retrieval system,
without permission in writing from the publishers.
Trademark notice: Product or corporate names may be trademarks
or registered trademarks, and are used only for identification
and explanation without intent to infringe.
British Library Cataloguing-in-Publication Data
A catalogue record for this book is available from the British Library
Library of Congress Cataloging-in-Publication Data
Names: Memmert, Daniel, author. | Raabe, Dominik, author.
Title: Data analytics in football : positional data collection, modelling and
analysis / Daniel Memmert and Dominik Raabe.
Description: Abingdon, Oxon ; New York, NY : Routledge, 2018. |
Includes bibliographical references and index.
Identifiers: LCCN 2017058008 (print) | LCCN 2018012668 (ebook) |
ISBN 9781351210164 (Master eBk) | ISBN 9781351210157 (Web PDF) |
ISBN 9781351210157 (ePub3) | ISBN 9781351210133 (Mobipocket/
Kindle) | ISBN 9780815381549 (hbk) | ISBN 9780815381556 (pbk) |
ISBN 9781351210164 (ebk)
Subjects: LCSH: Soccer—Data processing. | Soccer—Statistical methods. |
Soccer—Mathematical models. | Quantitative research.
Classification: LCC GV943 (ebook) | LCC GV943 .M385 2018 (print) |
DDC 796.33—dc23
LC record available at https://lccn.loc.gov/2017058008
Translation from the German language edition:
Revolution im Profifußball Mit Big Data zur Spielanalyse 4.0
by Daniel Memmert and Dominik Raabe
Copyright © Springer-Verlag GmbH Deutschland 2017
This Springer imprint is published by Springer Nature
The registered company is Springer-Verlag GmbH
ISBN: 978-0-8153-8154-9 (hbk)
ISBN: 978-0-8153-8155-6 (pbk)
ISBN: 978-1-351-21016-4 (ebk)
Typeset in Bembo and Stone Sans
by Florence Production Ltd, Stoodleigh, Devon, UK
CONTENTS
Preface by Hansi Flick vii
Preface by Ralf Rangnick ix
Acknowledgements xi
1 Where is the revolution? 1
2 A historical perspective on positional data 13
3 Technological background 35
4 Collecting data in the Bundesliga 43
5 In search of the Holy Grail 49
6 Betting and sports analytics 65
7 Where are they running? 71
8 From media to storytelling 77
9 Key properties of long-term success in football 85
10 The key to success 93
11 Reasons for dominance 99
vi Contents
12 FCB versus FCB 105
13 The myth of home advantage 123
14 Managerial influence 127
15 All on attack 133
16 Laws of a derby 137
17 Who will be nominated for the FIFA World Cup 2018? 147
18 Conclusion 153
Index 173
PREFACE
Hansi Flick
Across all levels and age groups there are a variety of factors that are responsible
for long-term fun and success for football teams. Among of these factors are tech-
nique, physical condition and cognition. In recent years especially, tactical aspects
have gained importance. The value of match plans increases—and they have to
cover more and more possible scenarios. As a consequence, coaches and assistant
coaches have to invest more work into the preparation to observe and evaluate all
aspects of the game. And this takes time, a lot of time.
Therefore a specific field of profession has developed: game analytics. This
offers a wide range of options in diagnosis and training control for the coaching
staff, not least because of the technical developments in this field. However,
“traditional” video analyses are time consuming even for these specialists, in
particular because complex game situations have to be determined and decoded
manually.
This book gives an overview of the genesis of the “game analysis.” In addition
to this, it shows that information about the complexity of the game can be
generated automatically and in seconds with the help of a new methodological
approach. Its basis is data that we have been collecting from Die Mannschaft
(German Men’s Football National Team) for more than ten years at the German
Football Association (Deutscher Fußball Bund, DFB): positional data! Now these
can be ennobled thanks to new technical developments.
In practice the value of positional data is mainly in regard to the control of
work load and physical stress. The integration of positional data in tactical analyses
has been not yet been utilized—although it would be possible to use these. At this
point, sport science and sports computer science, as well as the analysts, are equally
in high demand. Their task is to translate these data into a practical context: the
challenge is to not only use the technical possibilities to generate Big Data, but
also to purposively manage the development of players and teams.
viii Prefance by Hansi Flick
In the sportive mission statement of the DFB we rooted the game vision as our
fundament. Besides the individual game concept that each coach can shape
according to his or her team’s abilities, we have defined independent guidelines
in our concept of the game and put these at the forefront of the conception. These
form a kind of “quality characteristic” for the understanding of the game. According
to the idea of the game (to score goals) it is, for example, indispensable for us to
recognize and use spaces behind opponents when attacking.This applies to anyone
who plays football, no matter where and at what age.
It is important to be able to assess the skills based on the guidelines quickly,
reliably and objectively. And this requires the inclusion of positional data beyond
the video-based approach. Ultimately, we are talking about geometry and physics,
in other words cofigurations, directions, angles and speeds. All of this makes up
positional data. When we take advantage of this potential, we can completely
redefine and maybe even revolutionize in particular the spotting of talent.
Dear football fans, I hope you have a good time and gain new knowledge
through reading this great book.
Yours,
Hansi Flick
Former Sports director of the German Football Association
PREFACE
Ralf Rangnick
Today, modern match analysis must offer more than the bare evaluation of tackle
rates and distances covered, as it has become obvious that these are not, per se, the
difference between victory and defeat. Rather, it is a significantly more complex
analysis with more performance indicators that is going to provide new insights
into elite football. Big Data—or the so called “positional data”—help in recognizing
tactical patterns, as nowadays these are able to track the position of each player
and the ball very accurately. As a consequence, I have always put emphasis on the
importance of a competent and well-equipped analysis department for modern
match analysis, at all my previous coaching jobs as well as in my current profession
as sports director. So I constantly tried rapidly to adapt to new methodological and
digital trends to use these for our games during my time with VfB Stuttgart,
Hannover 96, 1899 Hoffenheim, Schalke 04 and recently in Leipzig.
The Institute of Training and Computer Science in Sport has been one of the
leading institutions in the development and testing of advanced key performance
indicators based on positional data. Therefore it is only consequent and logical that
the first book about Big Data in elite sports is presented by these authors. To apply
this knowledge systematically to our regular training and practices is going to be
the next step.
Dear readers, I wish you much pleasure in reading this book, and success and
joy with the world’s greatest pastime.
Yours,
Ralf Rangnick
Sports Director, RB Leipzig
ACKNOWLEDGEMENTS
At this point, we would like to thank our experts for their knowledge:
Ralf Rangnick (Sports Director, RB Leipzig)
Urs Siegenthaler (Head Scout, German National Team)
Prof. Martin Lames (TU Munich)
Lars O.D. Christensen (Talent Development and Training, FC Midtjylland)
Prof. Jürgen Perl (University of Mainz)
Tim Bagner (ChyronHego, Account Manager Deutsche Fußball-Liga (German
Football League))
Hansi Flick (Former Sports Director of the German Football Association)
Ulrich Forstner (National Coach ‘Science and Education’, German Hockey Asso-
ciation)
Dr Holger Broich (Head of Health and Fitness of Bayern Munich)
Ernst Tanner (Head of Youth Department, FC Red Bull Salzburg)
Dr Hendrik Weber (Head of Technology and Innovation, Deutsche Fußball-Liga
(German Football League))
Dr Daniel Link (TU Munich)
Prof. Arnold Baca (University of Vienna)
Dominik Meffert (German Sports University Cologne)
Prof. Dr Matthias Lochmann (University of Erlangen)
Chuck Korb (Senior Analyst, Boston Bruins)
Stefan Wagner (Global General Manager Sports & Entertainment, SAP SE)
Joachim Holzmüller (FIFA, Head of Football Technology Innovation)
1
WHERE IS THE REVOLUTION?
Big Data in professional sports
“We want a revolution,” says Christofer Clemens, head analyst of the Germany
national football team in an interview with the football magazine 11Freunde. And
he carries on: “We want to completely scrutinize match analysis.” The heart of this
revolution is constituted by the huge amount of data which high performance
football has been diligently accumulating for years now. Big Data has football under
control, just like in many other areas of our daily lives. It is about time to draw the
right conclusions from this treasure of information. Already by 2015, the Institute
of Training and Computer Science in Sport at the German Sport University in
Cologne had started the first post-professional Master’s degree, M.A. Game Analysis.
Central aims: nurturing innovation and creativity in working with game data.
It is only possible to innovatively work in practice and understand constant
changes as continuity with scientific know-how. The requirements of game
analysis as a professional field have risen since athletic demands have evolved, due
to increasing professionalization in sports games. Therefore, it involves finding
new paths for the analysis and interpretation of video and positional data.
In the future, match analysts will be involved more frequently in the athletic
performance of their teams. They will be involved in the development of game
ideas and the generation of solutions for specific problems in particular. The
demand of game analysis data is further growing with intensive usage in
broadcasting. Media professionals will have to be able to comprehend data in the
context of the game and understandably communicate these. Last, but not least,
it takes highly qualified specialists to develop analysis methods for evaluation,
analysis and presentation of analysis data on an advanced level.
As Ralf Rangnick mentioned in his Foreword, match analysis nowadays includes
more than the counting of tackles, received passes, or distance completed. These
2 Where is the revolution?
conventional parameters, collectively called event-data, provide quite concrete
insights into the world’s favorite sport. However, scientific studies demonstrate
that they hardly deliver any clues regarding the final outcome of a game. Yet,
guided by the large-scale introduction of positional data in professional football
during the last years, a completely new perspective has opened up which only
waits to be utilized.
The German Hockey federation, which has been promoting innovative game
analysis for several years, has also recognized this:
Take Bayern Munich in football: Recently, their opponents had it very
comfortable with Bayern’s constant passing game in which personal elements
did not play a role anymore. Every now and then we have also experienced
it like this: We constantly work but remain completely ineffective. The aim
has to be to unsettle the opponent much more frequently. At the end not
even the opponent analysts that evaluate everything and present it in 50 page
glossy brochures to their teams should know what will happen with the
Germans.
Valentin Altenburg, August 7, 2016, 4:45 p.m.;
source: ZeitOnline
This statement on the game analysis of tomorrow was given by the Germany
national hockey team coach Valentin Altenburg, just before the XXXIst Olympic
Games, 2016, in Brazil. Afterwards he won the bronze medal with his team. He
indirectly encouraged the fact that we need new impulses and innovations despite
the knowledge of quantitative and qualitative game analysis. How reliable are our
key performance indicators (KPIs)? Which interpretation is possible? What does
it mean for the training schedule?
One possibility of realization can be found within the term Big Data. Positional
data have been providing a new standard for a couple of years now, in order to
view teams’ variability and flexibility in a better way and extract constant patterns
more precisely. Advanced techniques nowadays enable us to detect the location
of every player on the field. Thereby, every action of the players on the field can
be registered manually or with the use of (semi)-automated methods.
In practice, various technical procedures show the position of all players in the
form of X–Y-coordinates—and at best in real time. The captured data are described
as positional or tracking data. Either special camera systems or mobile devices, which
the players wear under their shirts, are used for data collection. The subsequent
analysis based on these positional data can be generated in just a few seconds.
Whenever a team accomplishes a successful move, it is divided into its component
parts in the blink of an eye—including tactical details of offensive as well as
defensive behavior of their own and the opposing team.
Predominantly based on video data, the modern standard is that analysts and
coaches use the information gathered on physical, technical, and tactical player
and team performance to optimize training processes or for game preparation.
Where is the revolution? 3
However, objective performance acquisition using digital data is still not living up
to its full potential. The experiences of recent years have clearly shown that the
capability of theories and methods based on sport sciences, especially in the area
of positional data, does not yet cover the needs of high-performance sport.
Current research focuses on how robust findings in football and other team and
racket sports can be generated by means of modern procedures using computer
science and statistics. The heart of this is the highly complex, but for practical
issues indispensable, question of how the tactical and technical components of a
match can be analyzed in such a way that relevant conclusions for coaching
decisions can be made. Only in this way is it possible to optimally utilize the ever-
growing flood of data to increase competitiveness (Memmert et al., 2016a;
Memmert & Rein, 2018).
To get one step closer to this goal, sport scientists are constantly developing
and testing elaborate KPIs, which are supposed to deliver information on player
performance. Even though promising approaches can be detected within these,
there is still a visible deficit in the practical establishment of such performance
parameters. To date, only a small number of procedures and methods have brought
it close to marketability, as empirical testing is behind schedule. There is an
obvious lack of field studies in the professional realm that would complement
theory with practical insights (Memmert et al., 2016b).
FIGURE 1.1 FIFA World Cup Final 2006: Italy 6, France 4 (a.e.t.). To our
knowledge, this is one of the first games in which positional data were
measured (a. Zinedine Zidane (FRA) and Gennaro Gattuso (ITA);
b. positional data; c. graphic map of positional data)
Source: Eddy Lemaistre/Corbis Sport/Getty Images
continued . . .
Ball ;Buffon ;Grosso ;Materazzi ;Cannavaro ;Zambrotta ;Camoranesi;Perrota ;Pirlo ;De Rossi ;Gattuso
;To
52.04,33.13,0.00;10.00,36.00;34.00,22.97;28.97,30.00;28.20,39.00;29.97,50.00;50.97,60.88;51.00,12.00;40.00,31.00;–65000.00,–65000.00;40.00,36.02;51
52.04,33.13,0.00;10.00,36.00;34.00,22.97;28.97,30.00;28.25,39.00;29.97,50.00;50.97,60.93;51.00,12.00;40.00,31.00;–65000.00,–65000.00;40.00,36.02;51
52.04,33.13,0.00;10.00,36.00;34.00,22.97;28.97,30.00;28.27,39.00;29.97,50.00;50.97,60.93;51.00,12.00;40.00,31.00;–65000.00,–65000.00;40.00,36.02;51
52.04,33.13,0.00;10.00,36.00;34.00,22.96;28.97,30.00;28.30,39.00;29.97,50.00;50.97,60.94;51.00,12.00;40.00,31.00;–65000.00,–65000.00;40.00,36.04;51
52.04,33.13,0.00;10.00,36.00;34.00,22.96;28.97,30.00;28.34,39.00;29.97,50.00;50.97,60.96;51.02,12.00;40.00,31.00;–65000.00,–65000.00;40.00,36.04;51
52.04,33.13,0.00;10.00,36.00;34.00,22.95;28.97,30.00;28.38,39.00;29.97,50.00;50.97,61.00;51.02,12.00;40.00,31.00;–65000.00,–65000.00;40.00,36.05;51
52.04,33.13,0.00;10.00,36.00;34.00,22.95;28.97,30.00;28.43,39.00;29.97,50.00;50.97,61.02;51.02,12.00;40.00,31.00;–65000.00,–65000.00;40.00,36.06;51
52.04,33.13,0.00;10.00,36.00;34.00,22.94;28.97,30.00;28.46,39.00;29.97,50.00;50.97,61.02;51.02,12.00;40.00,31.00;–65000.00,–65000.00;40.00,36.09;51
52.04,33.13,0.00;10.00,36.00;34.00,22.93;28.97,30.00;28.51,39.00;29.97,50.00;50.97,61.03;51.03,12.00;40.00,30.97;–65000.00,–65000.00;40.00,36.11;51
52.04,33.13,0.00;10.00,36.00;34.00,22.93;28.97,30.00;28.54,39.00;29.97,50.00;50.97,61.05;51.05,12.00;40.00,30.97;–65000.00,–65000.00;40.00,36.13;51
52.03,33.25,0.00;10.00,36.00;34.00,22.92;28.97,30.00;28.59,39.00;29.97,50.00;50.97,61.09;51.06,12.00;40.00,30.97;–65000.00,–65000.00;40.00,36.14;51
52.02,33.48,0.05;10.00,36.00;34.00,22.89;28.97,30.00;28.62,39.00;29.97,50.00;50.97,61.10;51.09,12.00;40.00,30.97;–65000.00,–65000.00;40.00,36.18;51
52.02,33.52,0.09;10.00,36.00;34.00,22.87;28.97,30.00;28.67,39.00;29.97,50.00;50.97,61.12;51.11,12.00;40.00,30.96;–65000.00,–65000.00;40.00,36.20;51
52.02,33.65,0.12;10.00,36.00;34.00,22.85;28.97,30.00;28.69,38.97;29.97,50.00;50.97,61.12;51.12,12.00;40.00,30.96;–65000.00,–65000.00;40.00,36.25;51
52.00,33.79,0.14;10.00,36.00;34.00,22.85;28.97,30.00;28.72,38.97;29.97,50.00;50.97,61.18;51.14,12.00;40.00,30.95;–65000.00,–65000.00;40.00,36.29;51
52.00,33.91,0.15;10.00,36.00;34.00,22.81;28.97,30.00;28.76,38.97;29.97,50.02;50.97,61.10;51.18,12.01;40.00,30.95;–65000.00,–65000.00;40.00,36.30;51
52.00,34.04,0.15;10.00,36.00;34.00,22.78;28.97,30.00;28.78,38.97;29.97,50.02;50.97,61.25;51.20,12.02;40.00,30.94;–65000.00,–65000.00;40.00,36.36;51
51.98,34.18,0.14;10.00,36.00;34.00,22.76;28.97,30.00;28.81,38.96;29.97,50.02;50.97,61.27;51.25,12.05;40.00,30.93;–65000.00,–65000.00;40.00,36.38;51
51.98,34.31,0.12;10.00,36.00;34.00,22.72;28.97,30.00;28.84,38.96;29.97,50.02;50.97,61.30;51.27,12.05;40.00,30.92;–65000.00,–65000.00;40.00,36.43;51
51.97,34.45,0.09;10.00,36.00;34.00,22.68;28.97,30.00;28.86,38.95;29.97,50.03;50.97,61.34;51.30,12.06;40.00,30.92;–65000.00,–65000.00;40.00,36.46;51
51.96,34.57,0.05;10.00,36.00;34.00,22.67;28.97,30.00;28.88,38.95;29.97,50.05;50.97,61.38;51.36,12.09;40.00,30.89;–65000.00,–65000.00;40.00,36.52;51
51.97,34.66,0.00;10.00,36.00;34.00,22.61;28.97,30.00;28.89,38.95;29.97,50.06;50.97,61.43;51.39,12.10;40.00,30.87;–65000.00,–65000.00;40.00,36.54;51
51.97,34.57,0.00;10.00,36.00;34.00,22.59;28.97,30.00;28.93,38.95;29.97,50.09;50.97,61.46;51.43,12.13;40.00,30.85;–65000.00,–65000.00;40.00,36.59;51
51.97,34.61,0.00;10.00,36.00;34.00,22.53;28.97,30.00;28.93,38.93;29.97,50.11;50.97,61.52;51.47,12.14;40.00,30.84;–65000.00,–65000.00;40.00,36.63;51
51.97,34.57,0.00;10.00,36.00;34.00,22.51;28.97,30.00;28.93,38.90;29.97,50.12;50.97,61.52;51.52,12.17;40.00,30.81;–65000.00,–65000.00;40.00,36.65;51
51.97,34.54,0.00;10.00,36.00;33.97,22.45;28.97,30.00;28.94,38.89;29.97,50.15;50.97,61.59;51.59,12.21;40.00,30.78;–65000.00,–65000.00;40.00,36.70;51
51.97,34.52,0.00;10.00,36.00;33.97,22.43;28.97,30.00;28.95,38.87;29.97,50.15;50.97,61.62;51.61,12.25;40.00,30.76;–65000.00,–65000.00;40.00,36.72;51
51.97,34.45,0.00;10.00,36.00;33.97,22.36;28.97,30.00;28.96,38.86;29.97,50.21;50.97,61.65;51.68,12.27;40.00,30.72;–65000.00,–65000.00;40.00,36.77;51
51.97,34.42,0.00;10.00,36.00;33.97,22.34;28.97,30.00;28.96,38.84;29.97,50.25;50.97,61.69;51.71,12.30;40.00,30.68;–65000.00,–65000.00;40.02,36.79;51
FIGURE 1.1 Continued
Source: Eddy Lemaistre/Corbis Sport/Getty Images
16
10
17
15
10
4
7
2
14
1 22
8
13 8
11
13
17
9
6
7
25
24
FIGURE 1.1 Continued
Source: Eddy Lemaistre/Corbis Sport/Getty Images
6 Where is the revolution?
Video data
Match Analysis 1.0
(from 1950) Match Analysis 2.0
(from 1988) Match Analysis 3.0
Position data
Quantitative Evaluation
(Frequencies) (from 1996) Match
Qualitative Evaluation
(Expert assessment of Analysis 4.0
match situations) Physiological and
technical Evaluation
(from 2011)
(Pass routes, running Dynamic tactical
distance, sprints, evaluation (Patterns,
passes etc.) constellations,
interactions, complex
KPIs)
FIGURE 1.2 From game analysis 1.0 to game analysis 4.0
FIGURE 1.3 The digital revolution in elite football: game analysis provided in a matter
of seconds thanks to Big Data
The fundamentals have nevertheless been established, and the data era has
already grabbed football in its claws. On the following pages we will describe some
exciting results, from which game analysis 1.0–4.0 and in the future even 5.0 can
change and develop football (see Table 1.1). At first we will follow the development
of game analysis in general and match analysis based on positional data, in particular
from its infancy to the status quo. Subsequently, we take a look behind the curtain
regarding the techniques used, data material, situation in both the Bundesliga and
European football, as well as an outlook on the everyday life of game analysts in
other types of sport.
