Evaluation of player actions in football

Data-driven analysis of player performance based on event and positional data 

In a joint project with FC Augsburg, the University of Ulm is investigating how player performance in professional football can be assessed more comprehensively using modern data analysis methods. The focus is on the data-driven analysis of player actions and their contribution to the flow of the game.  In the public eye, player performance is often assessed on the basis of just a few key metrics – such as goals, assists or shots on target. However, a large part of the actual performance remains invisible. Actions such as running patterns, positioning, pressing behaviour, dribbling or passing contribute significantly to the course of the game, but are only inadequately taken into account in traditional statistics.  Against this backdrop, the project is developing a data-driven evaluation model that holistically analyses the contribution of individual player actions to the game system. This is based on extensive event data from football matches, which is evaluated using established methods such as ‘Valuing Actions by Estimating Probabilities’ (VAEP). The aim is to quantify the influence of individual actions on the probability of goals or conceded goals, thereby enabling a more nuanced assessment of player performance. Based on these analyses, interactive dashboards for visualising player performance and data-driven player profiles, amongst other things, are being developed. These enable the systematic identification of players’ strengths and weaknesses and can be utilised, for example, in scouting or strategic squad planning. Furthermore, an intelligent analysis agent is being developed that can generate data-driven recommendations for transfers or tactical adjustments, for instance. 

In the long term, the project pursues the vision of a holistic evaluation model that incorporates not only event data but also players’ positional data. This will enable aspects such as positioning or spatial control to be systematically analysed in future, in order to obtain as complete a picture as possible of player performance within the context of the game. 

Project partner: FC Augsburg

Research partner: Prof. Dr Oliver Müller, University of Paderborn

Project period: since 2025

Transfer

The analytical approaches developed in the project enable a data-driven and more objective assessment of player performance in professional soccer. This allows clubs to make more informed decisions in areas such as scouting, game analysis, and roster planning. Furthermore, the project contributes to the advancement of data-driven decision support in sports and demonstrates how methods from data science and AI can be applied in professional sports organizations.