Dermot Sheridan is a sports scientist and high-performance coach currently undertaking a PhD with Insight in DCU. Together with PhD supervisor Prof. Mark Roantree, he is examining athlete monitoring data collected by elite sports teams during game preparation. He writes about it below:
Teams monitor the training process to measure how their players are coping with the demands of training and games. One widely used method is Rating of Perceived Exertion (RPE) based on the Borg CR10 scale. This method allows players to rate both the physiological and psychological stress response that completing the session imposes on them. The goal is to analyse the relationship between RPE collected after the session and GPS (global positioning systems) captured during the session. If we can better understand this relationship, then it is possible to predict RPE using GPS metrics and use this information to plan our session more accurately to achieve the correct stimulus.
To provide some baseline metrics, we analysed data collected from an elite inter-county football team during the 2020 and 2021 seasons. For this research, we constructed three different Machine Learning models to predict RPE and then assessed each model’s performance using GPS metrics only. Our highest performing model demonstrated a 79% accuracy for RPE prediction. One of our findings was that using contextual information to enrich the data leads to greater accuracy. The level of performance shown by our models provides early evidence that prediction of RPE response to training load is feasible using GPS metrics.