Data-driven approach to defining football styles in major leagues

Abstract

This study proposes a data-driven methodology to define and compare styles of play in football, with a focus on the top four teams from the English, French, German, Italian, and Spanish leagues during the 2017/2018 season. Using event-based metrics derived from possession intervals, we constructed a feature matrix representing tactical behaviors at the match level. A Principal Component Analysis, followed by Varimax rotation, revealed four interpretable and distinct emergent playing styles. By projecting matches onto this style-based representation, we evaluated stylistic differences across leagues. A one-way Anova test confirmed significant inter-league variation in style prevalence. Furthermore, a random forest classifier successfully identified leagues based on the style representation, and a game-theoretic feature importance analysis uncovered consistent associations between specific styles and leagues. These findings provide a robust, reproducible framework for empirically analyzing football playing styles across competitive contexts.