Global Positioning System‐Derived Metrics and Machine Learning Models for Injury Prediction in Professional Rugby Union Players

In sports, injury prevention is a key factor for success. Although injuries are challenging to predict, new technologies and the application of data science can provide valuable insights. This study aimed to predict injury risk among professional rugby union players using machine learning (ML) models. We analyzed data from 63 professional rugby union players during three seasons, categorized them into forwards and backs, and further classified them into five specific positions (tight five, back row, scrum-half, inside backs, outside backs). The dataset included GPS data and derived metrics such as total workload in the 1, 2, and 3 weeks prior to injury, acute-to-chronic workload ratio over different time windows, monotony, and strain. Injury prediction was assessed separately for different player positions using five ML classification models: logistic regression, naïve Bayes (NB), support vector machine, random forest (RF), and eXtreme gradient boosting (XGBoost). RF performed best for forwards overall, with XGBoost excelling in the tight five and SVM in the back row, whereas among backs, RF led for inside backs and NB for outside backs. Additionally, feature importance plots were used to examine the impact of various factors on injury occurrence. In conclusion, our ML-based approach can effectively predict injuries, with average F1 scores up to 0.66 (± 0.14), particularly when applying a combination of GPS-derived metrics. Additionally, key characteristics indicative of injury for players in various positions have been successfully identified. These findings underscored the potential of ML to enhance injury prediction and inform tailored training strategies for athletes.