Related Experiment Video
Updated: Jul 29, 2026
06:52
An Inertial Measurement Unit Based Method to Estimate Hip and Knee Joint Kinematics in Team Sport Athletes on the Field
Published on: May 26, 2020
Using Machine Learning for Identification of Athlete Availability Predictors in a Multisport Elite Female Athlete
Sam R Moore1,2,3, Elena I Cantú1, Carly L Brantner4,5
1Department of Exercise and Sport Science, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina.
Journal of Strength and Conditioning Research
|July 7, 2026
Summary
Machine learning identified key athlete availability predictors in elite female athletes. Training load, recovery, and wellness were crucial, but team-specific models offered better insights than a general approach.
Area of Science:
- Sports Science
- Data Science
- Athlete Performance
Background:
- Athlete availability (AA) is critical for elite sports performance.
- Identifying predictors of AA can optimize training and reduce injury risk.
- Machine learning offers advanced analytical capabilities for complex sports data.
Purpose of the Study:
- To utilize machine learning (ML) to identify influential predictors of athlete availability (AA) in a cohort of elite female athletes.
- To compare the effectiveness of ML models across different sports (lacrosse and soccer) and at the individual team level.
- To determine the relative importance of training load, recovery, wellness, body composition, and force plate (FP) jump metrics in predicting AA.
Main Methods:
- Employed elastic net regression, a machine learning technique, to analyze data from 52 NCAA Division I female athletes (lacrosse and soccer).
- Collected data on training load, sleep recovery, wellness, body composition, and preseason FP testing throughout a competitive season.
- Assessed AA as the percentage of unmodified practices and games attended.
Main Results:
- Machine learning models successfully identified influential predictors of AA, with varying importance across training load, recovery, and wellness variables.
- Models demonstrated good fit, with root mean squared errors (RMSE) of 17.8% for the combined cohort, 8.9% for lacrosse, and 17.8% for soccer.
- Significant differences in predictors were observed between teams, with force plate data not being a predictor for lacrosse and body composition not for soccer, supporting single-team models.
Conclusions:
- Training load, recovery, and wellness are consistently important predictors of athlete availability across elite female sports.
- Individualized, single-team ML models provide a more accurate understanding of AA predictors compared to a generalized cohort model.
- Future research should focus on comprehensive modeling to proactively identify athletes at risk of low availability, refining data prioritization for practitioners.