Predicting Future Performance in Powerlifting: A Machine Learning Approach.

Area of Science:

• Sports Science
• Biomechanics
• Data Science

Background:

• Classic powerlifting has grown in popularity since 2012, increasing sports science research.
• Previous studies analyzed strength adaptations but lacked predictive models for performance.
• No models existed to predict classic powerlifting performance considering age, sex, and weight categories for training optimization.

Purpose of the Study:

• To develop and validate a machine learning linear regression model for predicting classic powerlifting performance.
• To incorporate individual characteristics like sex, age, weight, and training history into predictions.
• To generate European normative powerlifting data for talent identification and training enhancement.

Main Methods:

• A machine learning-based linear regression model was developed.
• The model utilized data including sex, age, weight, initial strength, and competition history.
• Validation involved comparing predicted performance against actual results.

Main Results:

• The dataset comprised 54,064 observations from 8,907 powerlifters.
• Normative data varied significantly across sex, age, and strength categories (p < 0.001).
• The model showed high accuracy (R² 0.90–0.94), strong correlations (r 0.95–0.97), and no significant bias.

Conclusions:

• The machine learning model accurately predicts individual powerlifting performance, considering personal attributes.
• It assists coaches and athletes in setting achievable training goals and tracking progress.
• Normative data stratified by demographics and strength levels provide valuable benchmarks.