Applications of Explainable Artificial Intelligence in Association Football: A Systematic Review
Artificial intelligence and machine-learning methods are increasingly applied in association football to analyse player and team performance, training and match demands, tactical behaviour, injury and health-related outcomes, video and movement data, and other sport-specific decision-making tasks. However, many high-performing models remain difficult for coaches, athletes, sport scientists, medical practitioners, and other stakeholders to understand or translate into practice. Explainable artificial intelligence (XAI) may address this limitation by identifying influential features, clarifying individual predictions, visualising model attention, and supporting more transparent and actionable decisions. This systematic review aims to identify, classify, and critically synthesise empirical applications of explainable or interpretable artificial intelligence in association football. Searches will be conducted in PubMed, Web of Science Core Collection, Embase, Scopus, SPORTDiscus via EBSCOhost, and IEEE Xplore Digital Library. The search strategy combines terms relating to association football, artificial intelligence and machine learning, and explainability or interpretability, including specific methods such as SHAP, LIME, feature importance, partial dependence, counterfactual explanations, attention-based explanations, saliency maps, Grad-CAM, decision trees, and rule-based models. Eligible studies will be original empirical investigations conducted in association-football contexts that apply an artificial intelligence, machine-learning, deep-learning, or data-driven modelling approach and provide an explicit explanation or interpretation output. Studies conducted exclusively in futsal or other non-association-football codes will be excluded. The review will examine application tasks, populations and data sources, model types, explainability methods, global and local explanation levels, model performance and validation, principal explanatory findings, practical actionability, and methodological limitations. Owing to the anticipated heterogeneity of study tasks, data, models, and outcomes, the findings are expected to be synthesised primarily through a structured narrative approach. The review will identify current evidence, methodological gaps, and priorities for developing transparent, valid, and practically useful AI systems in association football.