Artificial intelligence in football training: A systematic review

Abstract

Background: Although The integration of Artificial Intelligence into football training has expanded considerably. The existing literature predominantly aggregates match analytics, tactical modeling and injury prevention. No systematic review has specifically examined AI applications in training environments dedicated to the development of technical and tactical skills in football environment.Methods: In accordance with PRISMA guidelines, five electronic databases (Google Scholar, Scopus, PubMed, EBESCO and SPORTDiscus) were systematically searched. Eight empirical studies satisfied the inclusion criteria and were evaluated using the Mixed Methods Appraisal Tool. Results: Studies fell into two clusters: data-driven assessment and classification tools (n = 5) and AI-enhanced feedback interventions (n = 3). Machine learning and deep learning architecture, including Artificial Neural Networks (ANN) and Convolutional neural Networks (CNN), achieved high classification and skill discrimination accuracy in controlled settings. However real-world deployment revealed significant performance degradation: CNN shot detection sensitivity dropped from 75% in laboratory conditions to 45% in field scenarios, while Support Vector Machine (SVM) pass sensitivity reached only 20.9%. Among intervention studies, only one utilized a conventional coaching control group, demonstrating a 2.5fold greater improvement in tackling technique with AI driven 3D visual feedback (23.84%) compared to verbal coaching (9.63%). Conclusion: AI driven tools offer significant promise as objective, augmented coaching aids for movement analysis and skill feedback. The use of AI should be positioned as an augmented coaching aid rather than a replacement for coaching expertise, with adoption barriers being largely organizational rather than technical. Future research priorities include large scale RCTs, Longitudinal retention studies, standardized effect size reporting, and geographic diversification beyond the current Asia centric literature.