Contextualize player performance in football using Machine Learning techniques

Football practitioners such as managers, scouts and technical directors are faced with a multitude of challenges. Managers have to determine how to face the opponent team in the upcoming match, scouts have to make a call whether a possible transfer target would perform well at their club, and technical directors need to build their squad for the upcoming years. In order to make the best decisions, these football practitioners are increasingly using data to make the best-informed decision. In recent years, many football metrics have been developed to assist these practitioners in the choices they make including metrics that measure player performance by analyzing the actions that players perform. However, most of these metrics struggle to capture the context in which players perform their actions. This project will focus on developing methods that better capture the technical, tactical and mental aspects that affect player performance. These aspects include the players with whom the player is playing, the decisions that the players on the pitch make, the positioning of the defense and the circumstances in which the match is played. We will contextualize the performances of players in their matches using machine learning techniques applied to spatio-temporal tracking data and ball event data.