Football clubs are increasingly using large volumes of data that are collected during football matches to discover suitable reinforcements in the transfer market. Although several datadriven performance metrics have been proposed in the past few years, most existing metrics focus on the offensive performances of football players alone and disregard their defensive performances. To bridge this gap, we introduce a reliable metric to quantify football players’ defensive abilities. In particular, our metric measures each player’s ability to intercept the passes performed by the opposing team. We investigate the difficulty to perform of an interception in combination with the support of all teammates and, most exciting, we consider the missed opportunities to perform an interception as well. By involving statistical machine learning techniques, we gain insight into the defensive performances of players. According to pass, player and match specific factor characteristics, we clarify these performances. Moreover, we introduce a model to predict the probability of an interception and subsequently predict the probability of each player to perform an interception in two independent stages and one correlated step. To optimize our models, we evaluate the quality of each model’s predictions according to LogLoss and AUC-PR ranks the performances of the models. Our models outperform baseline models based on averages and domain knowledge. In line with our expectations, we reveal established names such as Van Dijk and Kant´e to provide outstanding defensive performances according to evaluations out of season 2018/2019. Moreover, we provide insight into young talented players who at the moment, lack the experience of performing interceptions, but are mostly well positioned.

Koning, A.J.
Erasmus School of Economics

Piersma, J.P.T. (2020, June 9). Valuing Defensive Performances of Football Players. Econometrie. Retrieved from