Constructing Football Club Ratings with League Strength Awareness Using a Modified Glicko Model

Shelopugin A.; Sirotkin A. · Lecture Notes in Computer Science (2026) · DOI: 10.1007/978-3-032-27272-0_4

Resumen

Evaluating football clubs across different leagues is a fundamental challenge in sports analytics, particularly when assessing performance transitions between competitions. We propose a league-aware club rating framework based on a modified Glicko system. Building on recent extensions of Glicko that incorporate order effects and strength-dependent draw probabilities, we adapt the framework specifically to football by introducing goal-scoring adjustments and league-aware rating initialization and transition dynamics. These modifications enable stable cross-league comparisons and more realistic modeling of promotion and relegation effects. In addition, we train a gradient boosting model for match outcome prediction using historical goal-based features. We further compare the proposed framework with matrix-based approaches, including the Dixon–Coles and Markov chain models. Experimental results demonstrate that the modified Glicko-based framework achieves higher predictive accuracy than competing methods while maintaining a transparent probabilistic structure and clear interpretability. The resulting ratings can be applied to player evaluation, transfer analysis, and match outcome forecasting across leagues. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.