Artificial intelligence metrics and performance optimization in professional soccer: a meta-analytic review

Abstract

Background and Study Aim. The integration of artificial intelligence (AI) and machine learning (ML) in professional soccer has transformed performance analysis. It enables objective quantification of offensive, defensive, and physical parameters. The aim of this meta-analytic review was to evaluate the effectiveness of AI-based metrics, including expected goals (xG), passing efficiency, injury prediction, defensive actions, and physical tracking, in optimizing soccer performance and decision-making.

Materials and Methods. Following PRISMA guidelines, studies published between 2000 and mid-2025 were retrieved from Scopus, PubMed, and Web of Science. Peer-reviewed empirical studies applying AI/ML models to professional soccer metrics were included. A random-effects meta-analysis was conducted to synthesize classification accuracy and effect sizes (Cohen’s d).

Results. From 280 identified records, 15 studies met qualitative inclusion criteria, and 10 provided quantitative data. Pooled effect sizes ranged from d = 0.47 to 0.64. The highest effects were observed for physical tracking (d = 0.64) and expected goals (d = 0.61*). Injury prediction, defensive actions, and passing accuracy also showed significant effects (p < .001). AI models achieved an average predictive accuracy of 88% and outperformed traditional analytical methods.

Conclusions. AI-driven performance metrics substantially enhance predictive accuracy, tactical evaluation, and injury prevention in soccer. Standardized datasets, explainable models, and longitudinal validation are essential for integrating AI into elite performance management.