Explore the vast range of titles available.

Daniel Link explores how data analytics can be used for studying performance in soccer. Based on spatiotemporal data from the German Bundesliga, the six individual studies in this book present innovative mathematical approaches for game analysis and player assessment. The findings can support coaches and analysts to improve performance of their athletes and inspire other researchers to advance the research field of sports analytics. Contents Individual Ball Possession in Soccer Real Time Quantification of Dangerousity A Topography of Free Kicks Match Importance Affects Activity Effect of Ambient Temperature on Pacing Depends on Skill Level Vanishing Spray Reduces Extent of Rule Violations Target Groups Lecturers and students of sports science, data analytics, computer science Experts in sports data, bookmakers, media companies, sports reporting, coaches and sports analysts The Author Dr. Daniel Link has been a lecturer and researcher at the Department of Sports and Health Sciences at the Technical University of Munich (TUM) since 2010. His research focuses on performance analysis in team sports, including technological aspects of data acquisition as well as the modeling of phenomena in sports. He supports top level teams and sport federations in implementing new approaches in match analysis.

The task of the 2017 Soccer Prediction Challenge was to use machine learning to predict the outcome of future soccer matches based on a data set describing the match outcomes of 216,743 past soccer matches. One of the goals of the Challenge was to gauge where the limits of predictability lie with this type of commonly available data. Another goal was to pose a real-world machine learning challenge with a fixed time line, involving the prediction of real future events. Here, we present two novel ideas for integrating soccer domain knowledge into the modeling process. Based on these ideas, we developed two new feature engineering methods for match outcome prediction, which we denote as recency feature extraction and rating feature learning. Using these methods, we constructed two learning sets from the Challenge data. The top-ranking model of the 2017 Soccer Prediction Challenge was our k-nearest neighbor model trained on the rating feature learning set. In further experiments, we could slightly improve on this performance with an ensemble of extreme gradient boosted trees (XGBoost). Our study suggests that a key factor in soccer match outcome prediction lies in the successful incorporation of domain knowledge into the machine learning modeling process.

How well can machine learning predict the outcome of a soccer game, given the most commonly and freely available match data? To help answer this question and to facilitate machine learning research in soccer, we have developed the Open International Soccer Database. Version v1.0 of the Database contains essential information from 216,743 league soccer matches from 52 leagues in 35 countries. The earliest entries in the Database are from the year 2000, which is when football leagues generally adopted the “three points for a win” rule. To demonstrate the use of the Database for machine learning research, we organized the 2017 Soccer Prediction Challenge. One of the goals of the Challenge was to estimate where the limits of predictability lie, given the type of match data contained in the Database. Another goal of the Challenge was to pose a real-world machine learning problem with a fixed time line and a genuine prediction task: to develop a predictive model from the Database and then to predict the outcome of the 206 future soccer matches taking place from 31 March 2017 to the end of the regular season. The Open International Soccer Database is released as an open science project, providing a valuable resource for soccer analysts and a unique benchmark for advanced machine learning methods. Here, we describe the Database and the 2017 Soccer Prediction Challenge and its results.

This paper introduces the Large Events Model (LEM) for soccer, a novel deep learning framework for generating and analyzing soccer matches. The framework can simulate games from a given game state, with its primary output being the ensuing probabilities and events from multiple simulations. These can provide insights into match dynamics and underlying mechanisms. We discuss the framework’s design, features, and methodologies, including model optimization, data processing, and evaluation techniques. The models within this framework are developed to predict specific aspects of soccer events, such as event type, success likelihood, and further details. In an applied context, we showcase the estimation of xP+, a metric estimating a player’s contribution to the team’s points earned. This work ultimately enhances the field of sports event prediction and practical applications and emphasizes the potential for this kind of method.

Elite soccer teams that participate in European competitions need to have players in the best physical and psychological status possible to play matches. As a consequence of congestive schedule, controlling the training load (TL) and thus the level of effort and fatigue of players to reach higher performances during the matches is therefore critical. Therefore, the aim of the current study was to provide the first report of seasonal internal and external training load that included Hooper Index (HI) scores in elite soccer players during an in-season period. Nineteen elite soccer players were sampled, using global position system to collect total distance, high-speed distance (HSD) and average speed (AvS). It was also collected session rating of perceived exertion (s-RPE) and HI scores during the daily training sessions throughout the 2015-2016 in-season period. Data were analysed across ten mesocycles (M: 1 to 10) and collected according to the number of days prior to a one-match week. Total daily distance covered was higher at the start (M1 and M3) compared to the final mesocycle (M10) of the season. M1 (5589m) reached a greater distance than M5 (4473m) (ES = 9.33 [12.70, 5.95]) and M10 (4545m) (ES = 9.84 [13.39, 6.29]). M3 (5691m) reached a greater distance than M5 (ES = 9.07 [12.36, 5.78]), M7 (ES = 6.13 [8.48, 3.79]) and M10 (ES = 9.37 [12.76, 5.98]). High-speed running distance was greater in M1 (227m), than M5 (92m) (ES = 27.95 [37.68, 18.22]) and M10 (138m) (ES = 8.46 [11.55, 5.37]). Interestingly, the s-RPE response was higher in M1 (331au) in comparison to the last mesocycle (M10, 239au). HI showed minor variations across mesocycles and in days prior to the match. Every day prior to a match, all internal and external TL variables expressed significant lower values to other days prior to a match (p<0.01). In general, there were no differences between player positions. Conclusions: Our results reveal that despite the existence of some significant differences between mesocycles, there were minor changes across the in-season period for the internal and external TL variables used. Furthermore, it was observed that MD-1 presented a reduction of external TL (regardless of mesocycle) while internal TL variables did not have the same record during in-season match-day-minus.

Using the modified Delphi method, we aimed to understand the attributes youth coaches and recruiters perceive as important when identifying skilled youth performance at the entry level of representative soccer in Australia (i.e., Under 13 years). Furthermore, we also aimed to describe the current methods youth coaches and recruiters use to assess and identify these attributes in youth players. Australian regional youth technical directors and coaches (n = 20) completed a three stage process, including an initial interview and two subsequent questionnaires, whereby attributes and qualities associated with talent identification were rated and justified according to the importance for youth player performance and talent identification. Results indicate a hierarchy of attributes recruiters perceive as important for Under 13 soccer performance, including technical (i.e., first touch, striking the ball, one-versus-one ability, and technical ability under pressure), tactical (i.e., decision-making ability) and psychological attributes (i.e., coachability and positive attitude). In addition, the findings indicated attributes and qualities not emphasised within the talent identification process including, physiological, anthropometrical, sociological and several psychological attributes. It is suggested talent recruiters apply a holistic multidisciplinary approach to talent identification, with the current findings potentially providing initial evidence to suggest recruiters do consider numerous attributes when selecting and identifying youth players.

Background Late-stage rehabilitation programs often incorporate ‘sport-specific’ demands, but may not optimally simulate the in-game volume or intensity of such activities as sprinting, cutting, jumping, and lateral movement. Objective The aim of this review was to characterize, quantify, and compare straight-line running and multi-directional demands during sport competition. Data Sources A systematic review of PubMed, CINAHL, SPORTDiscus, and Cochrane Central Register of Controlled Trials databases was conducted. Study Eligibility Criteria Studies that reported time-motion analysis data on straight-line running, accelerations/decelerations, activity changes, jumping, cutting, or lateral movement over the course of an entire competition in a multi-directional sport (soccer, basketball, lacrosse, handball, field hockey, futsal, volleyball) were included. Study Appraisal and Synthesis Methods Data was organized based on sport, age level, and sex and descriptive statistics of the frequency, intensity, time, and volume of the characteristics of running and multi-directional demands were extracted from each study. Results Eighty-one studies were included in the review ( n  = 47 soccer, n  = 11 basketball, n  = 9 handball, n  = 7 field hockey, n  = 3 futsal, n  = 4 volleyball). Variability of sport demand data was found across sports, sexes, and age levels. Specifically, soccer and field hockey demanded the most volume of running, while basketball required the highest ratio of high-intensity running to sprinting. Athletes change activity between 500 and 3000 times over the course of a competition, or once every 2–4 s. Studies of soccer reported the most frequent cutting (up to 800 per game), while studies of basketball reported the highest frequency of lateral movement (up to 450 per game). Basketball (42–56 per game), handball (up to 90 per game), and volleyball (up to 35 per game) were found to require the most jumping. Limitations These data may provide an incomplete view of an athlete’s straight-line running load, considering that only competition and not practice data was provided. Conclusions Considerable variability exists in the demands of straight-line running and multi-directional demands across sports, competition levels, and sexes, indicating the need for sports medicine clinicians to design future rehabilitation programs with improved specificity (including the type of activity and dosage) to these demands.

The paper describes Dolores, a model designed to predict football match outcomes in one country by observing football matches in multiple other countries. The model is a mixture of two methods: (a) dynamic ratings and (b) Hybrid Bayesian Networks. It was developed as part of the international special issue competition Machine Learning for Soccer. Unlike past academic literature which tends to focus on a single league or tournament, Dolores is trained with a single dataset that incorporates match outcomes, with missing data (as part of the challenge), from 52 football leagues from all over the world. The challenge involved using a single model to predict 206 future match outcomes from 26 different leagues, played from March 31 to April 9 in 2017. Dolores ranked 2nd in the competition with a predictive error 0.94% higher than the top and 116.78% lower than the bottom participants. The paper extends the assessment of the model in terms of profitability against published market odds. Given that the training dataset incorporates a number of challenges as part of the competition, the results suggest that the model generalised well over multiple leagues, divisions, and seasons. Furthermore, while detailed historical performance for each team helps to maximise predictive accuracy, Dolores provides empirical proof that a model can make a good prediction for a match outcome between teams x and y even when the prediction is derived from historical match data that neither x nor y participated in. While this agrees with past studies in football and other sports, this paper extends the empirical evidence to historical training data that does not just include match results from a single competition but contains results spanning different leagues and divisions from 35 different countries. This implies that we can still predict, for example, the outcome of English Premier League matches, based on training data from Japan, New Zealand, Mexico, South Africa, Russia, and other countries in addition to data from the English Premier league.

The presented field experiment in an 11 vs. 11 soccer game set-up is the first to examine the impact of different formations (e.g. 4-2-3-1 vs. 3-5-2) on tactical key performance indicators (KPIs) using positional data in a controlled experiment. The data were gathered using player tracking systems (1 Hz) in a standardized 11 vs. 11 soccer game. The KPIs were measured using dynamical positioning variables like Effective Playing Space, Player Length per Width ratio, Team Separateness, Space Control Gain, and Pressure Passing Efficiency. Within the experimental positional data analysis paradigm, neither of the team formations showed differences in Effective Playing Space, Team Separateness, or Space Control Gain. However, as a theory-based approach predicted, a 3-5-2 formation for the Player Length per Width ratio and Pressure Passing Efficiency exceeded the 4-2-3-1 formation. Practice task designs which manipulate team formations therefore significantly influence the emergent behavioral dynamics and need to be considered when planning and monitoring performance. Accordingly, an experimental positional data analysis paradigm is a useful approach to enable the development and validation of theory-oriented models in the area of performance analysis in sports games.

Participation in soccer match-play leads to acute and transient subjective, biochemical, metabolic and physical disturbances in players over subsequent hours and days. Inadequate time for rest and regeneration between matches can expose players to the risk of training and competing whilst not entirely recovered. In professional soccer, contemporary competitive schedules can require teams to compete in excess of 60 matches over the course of the season with periods of fixture congestion occurring, prompting much attention from researchers and practitioners to the monitoring of fatigue and readiness to play. A comprehensive body of research has investigated post-match acute and residual fatigue responses. Yet the relevance of the research for professional soccer contexts is debatable, notably in relation to the study populations and designs employed. Monitoring can indeed be invasive, expensive, time inefficient, and difficult to perform routinely and simultaneously in a large squad of regularly competing players. Uncertainty also exists regarding the meaningfulness and interpretation of changes in fatigue response values and their functional relevance, and practical applicability in the field. The real-world need and cost–benefit of monitoring must be carefully weighed up. In relation to professional soccer contexts, this opinion paper intends to (1) debate the need for post-match fatigue monitoring; (2) critique the real-world relevance of the current research literature; (3) discuss the practical burden relating to measurement tools and protocols, and the collection, interpretation and application of data in the field; and (4) propose future research perspectives.