Prediction of talent selection in elite male youth soccer across 7 seasons: A machine-learning approach.
Level 3 - non-randomized controlled study
Longitudinal observational cohort study evaluating predictive models for player selection
PubMed 39688281 · doi:10.1080/02640414.2024.2442850
What was done
Researchers tracked 409 elite male youth soccer players (generating 980 datapoints) across U12 to U19 age groups within a single professional German academy over a 7-year period. A multidimensional battery of physical, physiological, psychological, skill, health, age, and position-related variables was assessed. Supervised machine learning algorithms (XGBoost) were applied to evaluate variable importance and predict player selection versus deselection for the subsequent age group.
What was found
The XGBoost classification models achieved a receiver operating characteristic area under the curve (ROC-AUC) of 0.69 and an F1-score of 0.84. Physical and physiological parameters (linear sprint, change-of-direction sprint, countermovement jump, aerobic speed reserve) alongside soccer-specific skill tests showed the highest feature importance across age groups. Psychological metrics (motive structure, motive attention, motive competition, cognitive flexibility) were of medium predictive importance, whereas health-, age-, and position-related parameters showed no consistent pattern.
Why it matters
This study provides empirical, multi-season evidence demonstrating that objective physical, physiological, and sport-specific skill metrics predominantly drive retention decisions in elite academy pathways compared to psychological or contextual factors.
Limits
The study was conducted within a single German professional soccer academy, which may limit generalizability to other coaching philosophies or international developmental structures. Female players were not included. The discriminative capacity of the model was modest (ROC-AUC 0.69), indicating substantial unexplained variance in selection decisions.