Automated rowing event assignment: a machine learning approach.

Area of Science:

• Sports Science
• Biomechanics
• Machine Learning

Background:

• Elite rowing performance depends on matching athlete capabilities to event demands.
• Objective athlete classification is crucial for optimizing team composition and success.

Purpose of the Study:

• To develop and evaluate machine learning models for assigning rowers to specific rowing events.
• To utilize demographic and kinematic data for objective rower classification.

Main Methods:

• Collected 3D kinematic data (trunk, upper arm) from 55 elite rowers at three stroke rates using an inertia measurement unit system.
• Analyzed segmental and joint range of motion and motion coordination using vector coding.
• Trained six supervised machine learning models on demographic and kinematic data.

Main Results:

• Machine learning models successfully classified rowers into event groups (coxed eight vs. single/pair).
• Top models (decision tree, extreme gradient boosting, random forest) achieved high accuracy rates (0.89-0.93).

Conclusions:

• Automated rowing event assignment via machine learning provides coaches with objective decision-making tools.
• This approach minimizes subjective bias, improving the fairness and accuracy of athlete selection.
• Optimizing team composition through data-driven insights can enhance overall performance outcomes.