Overtraining syndrome (OTS) poses a critical challenge in youth soccer, particularly during periods of rapid physiological maturation combined with high training demands. This study aimed to develop and validate a multidimensional prediction model for overtraining risk in youth soccer players by integrating physiological, psychological, and performance parameters through advanced machine learning. A longitudinal study tracked 120 male youth players (aged 12-18) from six elite South Korean academies over one competitive season (August 2023-May 2024). Data included bi-weekly blood sampling (testosterone, cortisol, creatine kinase, IL-6, TNF-a), weekly psychological assessments (RESTQ-Sport, POMS), continuous GPS-based training load monitoring, and monthly performance tests. A random forest model with SMOTE to address class imbalance achieved an AUC-ROC of 0.94 (internal validation), with sensitivity and specificity of 0.87 and 0.92, respectively. Key predictors included testosterone-to-cortisol ratio (0.89), RESTQ-Sport balance (0.83), and acute:chronic workload ratio (0.78). A simplified, non-invasive model excluding blood markers achieved an AUC-ROC of 0.89. A three-tier risk stratification system identified 85% of high-risk cases a week before performance declined. These findings underscore the model`s superior predictive power and practical utility, offering a foundation for evidence-based, proactive overtraining risk management in elite youth soccer development.
© Copyright 2025 Journal of Sports Sciences. Taylor & Francis. All rights reserved.