The Kos Angle, an optimizing parameter for football expected goals (xG) models

Abstract

Abstract The utilization of metrics such as expected goals (xG) has the potential to provide teams with a competitive edge. By incorporating xG into their analysis and decision-making processes, teams can gain valuable insights. This study proposes a new approach to football xG modeling using Kos Angle which represents the shooting angle, from which we substract the angles occupied by players inside the shot angle. The objective of this study is to evaluate the impact of the Kos Angle feature on the performance of football xG models. After developing the mathematical formula of the Kos Angle, we selected additional features and built different xG models. Subsequently, the impact of the Kos Angle feature on the models’ performances was evaluated, revealing an increase in Recall and Precision and a decrease in Brier score and RMSE. We also found that the Kos Angle accounted for a significant portion of the models’ predictive power. By providing a more realistic representation of shot situations, the addition of the Kos Angle feature allows the improvement of xG models performances, which can give a more valuable insights to football professionals who rely on xG metrics and their variations.