Modeling Team-Compatibility Factors Using a Semi-Markov Decision Process: A Data-Driven Framework for Performance Analysis in Soccer

摘要

Player selection is one of the great challenges of professional soccer clubs. Despite extensive use of performance data, a large number of player transfers at the highest level of club soccer have less than satisfactory outcome. This paper proposes a Semi-Markov Decision Process framework to model the dependencies between players in a team and its effects on individual and team performance. The study uses data from the English Premier League to artificially replace players in their prospective teams and approximate the outcome using a simulation. A comparison between the expected and actual performance determines the goodness of the model. In early experiments, the output of the model has correctly identified the trend changes in scoring and conceding goals and provided a high correlation with actual performance.