Péter Sipos (LinkedIn)
This research examines whether a valid fatigue metric can be created based on publicly available event and injury data, and whether the predictive accuracy for Premier League matches can be improved by incorporating fatigue, form, and historical match data. The analysis, covering the 2015/16–2024/25 seasons, employs a RAPM model, ELO ratings, xG-based form, and per-minute workload metrics within an ordinal logistic regression and XGBoost framework. The results indicate that no reliable correlation can be established between workload metrics and injuries, and that the inclusion of contextualized variables does not provide any significant predictive advantage over the baseline model in either the player performance or match outcome models.