Models for Predicting the Success of a Footballer

Football statistical data can be used to drive important business decisions, especially when it comes to recruitment. With advances in data collection and Machine Learning techniques, such decision-based analyses are more prevalent and are providing increased accuracy. We investigate several traditional Machine Learning techniques for predicting the future performance of a footballer in order to determine whether to recruit them and at what price. These traditional approaches include XGBoost, Random Forest, K−Nearest Neighbours and the Support Vector Classifier. We also introduce a new Machine Learning model for performing this prediction and demonstrate the superiority (using the Mean Absolute Error metric) of this new approach. We believe that our work can be used by football scouts when looking at prospective players to make more precise offers to potential recruits.