A Supervised Learning Approach for Evaluating Football Performances

Over the past decades, the global sports market has experienced significant growth, with a compound annual growth rate (CAGR) of approximately 5% per year. The market value increased from $486.61 billion in 2022 to $512.14 billion in 2023, providing profitable opportunities for bookmakers and investors. In this context, the use of data is becoming increasingly significant, focusing on acquiring information about players and sporting events to predict outcomes and optimize team strategies. Therefore, the ability to accurately predict match outcomes is of growing research interest. This study addresses the crucial challenge of predicting the number of goals scored in football matches of the main Italian league (Serie A) based on data from previous seasons to investigate which approach can significantly improve the accuracy of predictions and contribute to the understanding of the determining factors behind the results of football matches. The research employs specialized regression models, such as machine learning methods, generalized linear models and some of their extensions designed to handle count-type response variables and account for overdispersion and excess zeros commonly found in goal-scoring patterns. By utilizing these models, we aim to improve prediction accuracy and facilitate informed decision-making for sports fans and betting professionals.