Expected Goals in Soccer: Explaining Match Results using Predictive Analytics

Where data based decision making is taking over businesses and elite sports, elite soccer is lacking behind. In elite soccer, decisions are still often based on emotions and recent results. As results are, however, dependent on many aspects, the reasons for these results are currently unknown by the elite soccer clubs. In our study, a method is proposed to determine the expected winner of a match. Since goals are rare in soccer, goal scoring opportunities are analyzed instead. By analyzing which team created the best goal scoring opportunities, a feeling can be created which team should have won the game. Therefore, it is important that the quality of goal scoring opportunities accurately reflect reality. Therefore, the proposed method ensures that the quality of a goal scoring opportunity is given as the probability of the goal scoring opportunity resulting in a goal. It is shown that these scores accurately match reality. The quality scores of individual goal scoring opportunities are then aggregated to obtain an expected match outcome, which results in an expected winner. In little more than 50% of the cases, our method is able to determine the correct winner of a match. The majority of incorrect classified winners comes from close matches where a draw is predicted. The quality scores of the proposed method can already be used by elite soccer clubs. First of all, these clubs can evaluate periods of time more objectively. Secondly, individual matches can be evaluated to evaluate the importance of major events during a match e.g. substitutions. Finally, the quality metrics can be used to determine the performance of players over time which can be used to adjust training programs or to perform player acquisition. Expected Goals in Soccer: Explaining Match Results using Predictive Analytics iii