Return to performance: machine learning insights into how absence time following muscle injuries affects match running performance in LaLiga soccer players.

Abstract

To determine how absence time after muscle injuries affects external load metrics in elite soccer players and identify which performance variables are most impacted by the injury. A total of 110 lower limb muscle injuries from LaLiga players were analysed. Following an analysis of pre- and post-injury data to identify which outcomes were affected by muscle injury, machine learning algorithms were employed to examine relationships between absence duration and performance metrics. Maximal speed, maximal acceleration, maximal deceleration, composite index (i.e., overall player performance) and sprint count during matches were the most affected variables after return to play. The multiple linear regression (MLR) model and random forest regression (RFR) presented an R² of 0.348 and 0.442. Maximal speed was the variable most strongly associated with absence time in both models (coefficient in MLR = 7.94; mean absolute SHAP value in RFR model = 4.99), with longer recovery periods correlating with reduced match performance in this metric. Maximal acceleration and deceleration also showed declines with increased absence time. In contrast, sprint count exhibited no significant relationship with absence time. Maximal speed, acceleration and deceleration capacity, as well as sprint count and overall performance, are affected after muscle injuries. However, prolonged recovery following muscle injuries especially reduces maximum speed and acceleration/deceleration capacity in elite players during matches, while sprinting actions remain unaffected by absence time.