Critical speed models of high-resolution speed-duration profiles describe peak running demands in soccer

Abstract

The purpose of this study was to characterize critical speed (CS) models for summarizing high-resolution speed-duration profiles from GPS tracking data obtained from soccer players. GPS data from 15 male NCAA Division I soccer players were collected during practices and games over a 6-week period. Moving averages of the speed data were computed for each file for duration windows spanning 0.1 to 600 seconds at 0.1-second resolution. Speed-duration profiles for each session and for the entire sampling period (“global”) were generated for each player by selecting the maximal mean speeds for each duration. Four models were fit to the profiles: the two-parameter CS (CS2) model, the three-parameter CS (CS3) model, the omni-domain speed-duration (OmSD) model, and the five-parameter logistic (5PL) model. The 5PL, CS3, and OmSD models exhibited similar goodness of fits, and all outperformed the CS2 model. Similar CS estimates were obtained for each model, whereas maximum speed ( Smax) estimates were lower for OmSD compared to the 5PL. Players exhibited a range of parameter values for CS, D′, and Smax. Smax and CS estimated from session-specific speed-duration profiles were on average higher for games compared to practices. We conclude that CS models are useful for empirically describing speed-duration profiles and for assessing peak running demands for soccer practices and games. The proposed approach could help coaches design practice activities to better mimic game demands.