Interpretable Machine Learning for Football Performance Analysis: Evidence of Limited Transferability from Elite Leagues to University Competition

TL;DR

This study examines whether interpretability of machine learning models in football performance analysis remains reliable when transferring from elite leagues to university-level competitions, revealing significant domain-dependent instability.

Contribution

It demonstrates that interpretability methods may not be robust across different competition levels, highlighting the need for domain-specific validation in sports analytics.

Findings

Elite football determinants are stable across leagues and models.

University football shows reordering of key indicators and explanation instability.

Interpretability robustness varies significantly under domain shift.

Abstract

Machine learning has become increasingly prevalent in football performance analysis, yet most studies prioritize predictive accuracy while implicitly assuming that learned performance determinants and their interpretations are transferable across competition levels. Whether interpretability remains reliable under domain shift-from elite to university football remains largely unexplored. This study investigates whether performance determinants learned from elite competitions are structurally transferable to university-level football and whether their interpretations remain robust under domain shift. Models were trained on large-scale event data from the top five European leagues and applied to university football data from National Tsing Hua University (NTHU) using an identical feature space. Random Forest and Multilayer Perceptron models were interpreted using SHapley Additive exPlanations (SHAP) and Counterfactual Impact Score (CIS). Across five experiments, elite football exhibited a stable and consistent hierarchy of performance determinants across leagues, models, and explanation methods. In contrast, NTHU university football showed substantial reordering of key indicators, reduced explanation stability, weaker structural agreement with elite domains, and increased sensitivity to explanation method. These findings suggest that interpretability robustness is domain-dependent. Rather than reflecting methodological limitations alone, instability in explanations under domain shift may serve as a diagnostic signal of structural ambiguity in the target domain.