Workload, maturation, and injury risk in youth football: A machine learning approach
Abdulkareem Babatunde Taiwo, RAFIU OLAOLUWA Okuneye, Fatai Akinola Apalara, Halimat I’ya Ismail-Orire, Okechukwu Kingsley Oforka, Michael Ayodele Adefuye et al. · Team Sports Studies · 2026
Background: Injury risk in youth football is influenced by training load and biological maturation, yet the combined predictive value of these factors remains unclear.
Objective: To determine whether integrating external load, internal load, and maturation variables improves short-term injury prediction in youth football players.
Methods: A prospective cohort study followed 240 male academy football players aged 12–18 years across three professional academies over 12 months. External load was assessed using GPS-derived metrics, internal load using session rating of perceived exertion (sRPE), and maturation using maturity offset relative to peak height velocity. The outcome was the occurrence, within the subsequent seven days, of a time-loss injury resulting in at least one missed training session or match. Random forest and LASSO logistic regression models were evaluated using nested temporal cross-validation.
Results: The load and maturity random forest model demonstrated the best predictive performance (AUC = 0.78, 95% CI 0.74–0.82), outperforming baseline demographic models (AUC = 0.62, 95% CI 0.57–0.67) and load-only models (AUC = 0.76, 95% CI 0.71–0.80). Acute: chronic workload ratio, maturity offset, and weekly high-speed distance were the strongest predictors.
Findings: remained consistent across subgroups and sensitivity analyses.
Conclusion: Combining workload and maturation variables improved short-term injury prediction in youth football compared with demographic-only models. Further external validation is required before routine implementation in academy decision-making.