Using Multilevel Models to Compare Performance Prediction and Characterization Abilities Between Power-Law and...

Area of Science:

• Sports Science
• Exercise Physiology
• Biomechanical Analysis

Background:

• Speed-duration relationships are crucial for predicting athletic performance and characterizing runners.
• The critical-speed (CS) and power-law (PL) models are commonly used by practitioners.
• A comparative analysis using multilevel models (MLMs) is needed to evaluate these models.

Purpose of the Study:

• To compare the performance prediction accuracy of the PL and CS models.
• To assess the athlete characterization capabilities of the PL and CS models.
• To utilize multilevel models (MLMs) for middle- and long-distance runners.

Main Methods:

• Analysis of 184,755 performances from 52,847 athletes across 6 running events (400m-10,000m).
• Development of MLMs accounting for the hierarchical data structure.
• Evaluation of prediction accuracy using mean absolute error (MAE) and mean absolute relative error (MARE).
• Athlete characterization via model-derived parameters: speed (S), endurance (E), anaerobic capacity (D'), and CS.

Main Results:

• The PL MLM demonstrated significantly higher prediction accuracy (MAREtime = 1.32%, MAREspeed = 1.32%) than the CS MLM (MAREtime = 6.87%, MAREspeed = 9.1%).
• Long-distance runners showed higher E and CS; middle-distance runners exhibited higher S.
• Specialization, gender, and performance level influenced model-derived parameters (S, E, D', CS).

Conclusions:

• MLMs provide enhanced understanding of speed-duration relationships across athlete clusters, gender, and performance levels.
• Model-derived parameters offer valuable insights for comprehensive athlete profiling.
• The PL MLM is identified as a reliable and accurate tool for predicting running performance.