V _max estimate from three-parameter critical velocity models: validity and impact on 800 m running performance prediction

Abstract

The purpose of this study was to evaluate the validity of maximal velocity (V _max) estimated from three-parameter systems models, and to compare the predictive value of two- and three-parameter models for the 800 m. Seventeen trained male subjects $({\text{\ifmmode\expandafter\dot\else\expandafter\.\fi{V}O}}_{2} \max = 66.54 \pm 7.29\,{\text{ml}}\,{\text{min}} ^{{ - 1}} \,{\text{kg}}^{{ - 1}} )$ performed five randomly ordered constant velocity tests (CVT), a maximal velocity test (mean velocity over the last 10 m portion of a 40 m sprint) and a 800 m time trial (V _800 m). Five systems models (two three-parameter and three two-parameter) were used to compute V _max (three-parameter models), critical velocity (CV), anaerobic running capacity (ARC) and V _800 m from times to exhaustion during CVT. V _max estimates were significantly lower than (0.19<Bias<0.24 m s⁻¹) and poorly associated (0.44<r<0.49) with actual V _max (8.43±0.33 m s⁻¹). Critical velocity (CV) alone explained 40–62% of the variance in V _800 m. Combining CV with other parameters of each model to produce a calculated V _800 m resulted in a clear improvement of this relationship (0.83<r<0.94). Three-parameter models had a better association (0.93<r<0.94) and a lower bias (0.00<Bias<0.04 m s⁻¹) with actual V _800 m (5.87±0.49 m s⁻¹) than two-parameter models (0.83<r<0.91, 0.06<Bias<0.20). If three-parameter models appear to have a better predictive value for short duration events such as the 800 m, the fact the V _max is not associated with the ability it is supposed to reflect suggests that they are more empirical than systems models.