A0929
Title: Parametric and semiparametric mixed modelling of athletic ability in young soccer players, considering injuries
Authors: Arthur Guillotel - Stade Rennais Football Club (France)
Brigitte Gelein - ENSAI (France) [presenting]
Benoit Bideau - Rennes 2 University (France)
Anthony Sorel - Rennes 2 University (France)
Abstract: Over four consecutive seasons, a professional soccer academy was closely monitored, with data collected on key athletic performance metrics such as speed, strength, power, and endurance. To model how athletic abilities evolve with age, both parametric and semi-parametric mixed models are applied. In the semi-parametric approach, the fixed effects were estimated using machine learning methods, specifically gradient boosting and random forests. These techniques allow for greater flexibility by capturing non-linear relationships between predictors and outcomes. Importantly, the models accounted not only for age but also for injury history, a critical yet often neglected factor in athletic development. Several injury burden indicators that reflect both the duration and severity of injuries are constructed and compared in different ways. The results highlight the value of semiparametric mixed modeling approaches in producing accurate and robust predictions, with low error margins across most athletic parameters. Furthermore, the models reveal distinct developmental trajectories influenced by a combination of age and injury history. These findings provide meaningful insights to support performance monitoring, training optimization, and the long-term development of young elite soccer players.
