VARIATION IN PERFORMANCE OF ELITE CYCLISTS FROM RACE TO RACE

Abstract

A performance test is useful for tracking the smallest worthwhile changes in an athlete's performance only if the typical (standard) error of measurement of performance in repeated tests is no greater than an athlete's typical variation of performance in repeated competitions. Typical errors of measurement in some tests are known; typical variations in races of most sports are not. We have therefore used repeated-measures analysis of log-transformed race times of elite male cyclists in various race series to derive typical variations of race times as coefficients of variation (CV). The series were: Kilo time trials in velodromes (2 seasons each of 3 races of 1 km over 15 wk); individual road time trials (3 races of 46–75 km over 27 d, including the Sydney Olympic race); World Cup road races (10 races of 230–294 km over 29 wk); and the Tour de France (2 individual time trials of 17 and 59 km, and 18 road-race stages of 145–255 km over 23 d). We performed separate analyses in each series for subgroups of cyclists ranked by mean speed. CV (and 95% likely limits) of cyclists in the fastest subgroups were: Kilo, 0.5% (0.3–1.3%); road time trials, 1.3% (0.9–2.5%); World Cup road races, 0.4% (0.3–0.5%); Tour de France time trials, 1.7% (1.3–2.7%); and Tour de France road races, 0.4% (0.8–1.0%). The CV in road races were lower than CV in time trials of comparable duration, owning to the averaging effect of cyclists riding in packs in the road races. The lower CV in the kilo relative to CV in longer time trials may be due to variability in pace chosen by the cyclist or variability in wind speed in the longer time trials. The CV for performance time in the best lab and field tests of cycling performance are similar to or smaller than these CV, so such tests should be suitable for tracking the competitive performance of elite cyclists.