Forecasting Goal Performance for Top League Football Players: A Comparative Study

Abstract

In this paper, we review the literature on Sports Analytics (SA) and predict football players’ scoring performance. Based on previous years’ performance, we predict the number of goals that players scored during the 2021–22 season. To achieve this, we collected advanced statistics for players from five major European Leagues: the English Premier League, the Spanish La Liga, the German Bundesliga, the French Ligue1 and the Italian Serie A, for seasons from 2017–18 up to 2021–22. Additionally, we used one-season lag features, and three supervised Machine Learning (ML) algorithms for experimental benchmarking: Linear Regression (LR), Random Forest (RF) and Multilayer Perceptron (MLP). Furthermore, we compared these models based on their performance. All models’ results are auspicious and comparable to each other. LR was the best performing model with Mean Absolute Error (MAE) 1.60, Mean Squared Error (MSE) 7.06 and Root Mean Square Error (RMSE) 2.66. Based on feature importance analysis, we established that every player’s upcoming scoring performance is strongly associated with previous season’s goals (Gls) and expected goals (xG).