Sex-specific pacing strategies across performance levels in 2000 m indoor rowing: Explainable machine learning evidence for coaching practice

This study investigated sex-specific pacing strategies across performance levels in 2000 m indoor rowing and examined how explainable machine learning can derive model-based tactical indicators for coaching practice. A total of 3125 official race records from the World Indoor Rowing Championships (2021–2025) were analysed. Sex-specific eXtreme Gradient Boosting models with SHapley Additive exPlanations (SHAP) were used to quantify the contribution of pacing-related variables to model-predicted 2000 m performance. The models showed acceptable predictive performance, with an overall root mean square error of 41.16 s. Rather than developing a comprehensive physiological prediction model, the analysis focused on race-derived pacing variables that are directly observable in coaching contexts. Elite athletes exhibited lower pacing variability than non-elite athletes ( p  < .001). Results also showed distinct model-derived tactical patterns between male and female elite athletes. Male elite rowers showed greater model-derived contribution from stroke-rate maintenance and final-segment pacing, whereas female elite rowers showed model-derived patterns consistent with lower tolerance for pacing variability and greater importance of precise pacing during the first 1000 m. As performance level increased, the model-derived importance of stroke rate decreased, while pacing-related variables showed greater contribution, suggesting greater relevance of pacing strategy at higher performance levels. SHAP analysis further identified exploratory, model-derived transition ranges, including approximately 33 spm for males and 24.75% for first-segment pacing in females. These findings show how explainable machine learning can translate large-scale competition records into interpretable performance indicators, providing coaches with exploratory, sex-specific reference points for monitoring pacing execution and race planning.