Abstract
This study investigated the relationship between acute and chronic training load metrics and non-contact muscle injuries in elite soccer players employing a novel statistical approach. A retrospective analysis was conducted during the 2020/21 season on 30 senior outfield players from an English Premier League club. Global Positioning System (GPS) technology monitored total distance, high-speed running (HSR) distance (5.5-7 m/s), sprint distance (> 7 m/s), and peak speed during training sessions and matches. A total of 42 injuries were documented, with an incidence of 8.94 injuries per 1000 hours, although only 12 non-contact muscle injuries were included in the analysis that occurred at 2.5 per 1000 hours of exposure. Acute (7-day) and chronic (28-day) training loads were examined, and data preprocessing addressed missing values and multicollinearity. To address class imbalance, the dataset was balanced using the Synthetic Minority Over-Sampling Technique (SMOTE) prior to logistic regression. Four significant predictors were retained: acute HSR (β = -0.175, p < 0.001), acute sprint distance (β = -0.613, p < 0.001), acute peak speed (β = 1.101, p < 0.001), and chronic total distance (β = 2.234, p < 0.001). The model demonstrated excellent discriminative ability with an AUC-ROC of 0.80. The results showed that higher acute volumes of HSR and sprint distance serve as protective factors against non-contact muscle injuries, whereas an increase in acute peak speed and chronic total distance significantly elevates injury risk. These findings underscore the importance of regular exposure to HSR to enhance injury resilience, while excessive load and peak speed may contribute to neuromuscular fatigue and overload.