Predicting Soccer Players' Fitness Status Through a Machine-Learning Approach

Area of Science:

• Sports Science
• Machine Learning in Athletics
• Soccer Performance Analysis

Background:

• Assessing soccer player fitness is crucial for performance and injury prevention.
• Traditional fitness tests may not fully capture in-training physiological responses.
• Machine learning offers novel approaches to monitor athlete status.

Purpose of the Study:

• Develop a machine-learning index to predict soccer player fitness.
• Validate the index against submaximal run tests (SMFT).
• Analyze training load's impact on the index and SMFT outcomes.

Main Methods:

• Collected training load data (external and internal) from 50 professional soccer players.
• Utilized machine learning to predict heart rate responses during training.
• Calculated a fitness index based on actual vs. predicted heart rates.
• Correlated the fitness index with SMFT results.

Main Results:

• Random forest regression was the best-performing machine learning algorithm.
• Key predictors for the fitness index included average speed, training duration, and work:rest ratio.
• The fitness index showed a strong correlation (r = .70) with SMFT outcomes.
• Divergence between the index and SMFT over the season indicated different fitness aspects.

Conclusions:

• Introduced an "invisible monitoring" method for in-training fitness assessment in soccer.
• The fitness index complements traditional tests for a holistic view of player readiness.
• Enables personalized training adjustments and supports injury prevention strategies.