Precision Training Via Causal Machine Learning: Modeling Rating of Perceived Exertion in Professional Soccer Players

Area of Science:

• Sports Science
• Machine Learning
• Data Analytics

Background:

• Managing training loads is crucial for optimizing performance and preventing injuries in professional soccer.
• The rating of perceived exertion (RPE) is a key metric for monitoring player fatigue and training stress.

Purpose of the Study:

• To explore predictive and prescriptive machine learning models for managing training loads in professional soccer.
• To evaluate the effectiveness of these models in predicting and prescribing optimal training regimens based on RPE.

Main Methods:

• Analysis of data from 14 professional soccer players over a full competitive season.
• Comparison of predictive models (linear regression, random forest, XGBoost) using RMSE and MAE.
• Utilized SHapley Additive exPlanations for feature importance and developed a counterfactual recurrent network for prescriptive modeling.

Main Results:

• The XGBoost model achieved the best predictive performance for RPE (RMSE: 1.262), identifying session distance as a key driver.
• The prescriptive counterfactual recurrent network model (RMSE: 1.379) enabled simulation of future RPE trajectories under varied training scenarios.

Conclusions:

• Predictive modeling accurately estimates RPE, while prescriptive modeling optimizes training strategies for personalized prescription.
• Integrating these AI approaches supports data-driven decision-making in soccer, enhancing player performance and recovery.
• Future research should focus on larger sample sizes and validation across different sports.