Player Tracking Data Analytics as a Tool for Physical Performance Management in Football: A Case Study from Chelsea Football Club Academy

Abstract

Background: Global positioning system (GPS) based player movement tracking data are widely used by professional football (soccer) clubs and academies to provide insight into activity demands during training and competitive matches. However, the use of movement tracking data to inform the design of training programmes is still an open research question. Objectives: The objective of this study is to analyse player tracking data to understand activity level differences between training and match sessions, with respect to different playing positions. Methods: This study analyses the per-session summary of historical movement data collected through GPS tracking to profile high-speed running activity as well as distance covered during training sessions as a whole and competitive matches. We utilise 20,913 data points collected from 53 football players aged between 18 and 23 at an elite football academy across four full seasons (2014–2018). Through ANOVA analysis and probability distribution analysis, we compare the activity demands, measured by the number of high-speed runs, the amount of high-speed distance, and distance covered by players in key playing positions, such as Central Midfielders, Full Backs, and Centre Forwards. Results and Implications: While there are significant positional differences in physical activity demands during competitive matches, the physical activity levels during training sessions do not show positional variations. In matches, the Centre Forwards face the highest demand for High Speed Runs (HSRs), compared to Central Midfielders and Full Backs. However, on average the Central Midfielders tend to cover more distance than Centre Forwards and Full Backs. An increase in high-speed work demand in matches and training over the past four seasons, also shown by a gradual change in the extreme values of high-speed running activity, was also found. This large-scale, longitudinal study makes an important contribution to the literature, providing novel insights from an elite performance environment about the relationship between player activity levels during training and match play, and how these vary by playing position.

Artificial intelligence for team sports: a survey

“…Price et al (2004) highlight the nature and severity of injuries that occur at academy level, and le Gall et al (2010) evaluate the fitness characteristics of young players in youth academies, highlighting which of these characteristics improve players chances of proceeding to higher levels. De Silva et al (2018) have also used the player tracking data that are available as a tool for training youth players and for physical performance management in football. They tested their work in a professional Premier League football academy.…”

PlayeRank

“…Our survey consisted of a set of pairs of players randomly generated by a two-step procedure, defined as follows: First, we randomly selected 35% of the players in the dataset. Second, for each selected player u, we cyclically iterated over the ranges [1,10], [11,20], and [21, ∞] and selected one value, say x, for each of these ranges, and then picked the player being x positions above u and the one being x positions below u in the role-based ranking (if they exist). This generated a set P of 211 pairs involving 202 distinct players.…”

“…We found that c maj = 68% and c una = 74%, indicating that PlayeRank has in general a good agreement with the soccer scouts, compared to the random choice (for which c maj = c una = 50%). Figure 11 offers a more detailed view on the results of the survey by specializing c maj and c una on the three ranges of ranking differences: [1,10], [11,20], [21, ∞]. The bars show a clear and strong correlation between the concordance among scouts' evaluations (per majority or unanimity) and the difference between the positions in the ranking of the checked pairs of players: When the ranking difference is ≤10 it is c maj = 59% and c una = 61%; for larger and larger ranking differences, PlayeRank achieves a much higher concordance with experts, which is up to c maj = 86% and c una = 91% when the ranking difference is ≥20.…”

“…Clearly, the disagreement between PlayeRank with the soccer scouts is less significant when the players are close in the ranking (i.e., their distance <10). Indeed, the comparison between soccer players is a well-known difficult problem, as witnessed by the significant increase in the fraction of unanimous answers by the scouts, which goes from a low 58% in the range [1,10] to a reasonable 71% in the range [21, +∞]. This a fortiori highlights the robustness of PlayeRank: The scouts' disagreement decreases as pairs of players are farther and farther in the ranking provided by PlayeRank.…”

Editorial: Fatigue and Recovery in Football

“…Finally, the study by De Silva et al [9] is the first to investigate the differences in activity demands during training and competitive matches in relation to playing positions in soccer at an elite soccer academy in order to understand differences between training and match sessions, with respect to different playing positions. The results indicate that, while there are significant position-specific differences in activity levels during matches, such differences are not observed for data pertaining to the training sessions.…”