Methods for tracking athletes' competitive performance in skeleton

In many sports, changes in performance time between races arising from differences in venues and weather far exceed changes in an athlete's true ability. Here we compare three methods to track performance of individual athletes in one such sport, skeleton. We developed the methods with official times of 33 male and 34 female athletes competing in three or more of 26 World Cup races over 4 years leading up to, but not including, the 2006 Winter Olympics. For two methods accessible to coaches, we fitted simple quadratic trajectories to each athlete's race placing and to percent behind the winning time. For a more sensitive method, we fitted similar quadratic trajectories to race time using a mixed model to adjust for mean race times. Correlations between predicted and observed performance in the races used to develop the methods were all similar (0.7). Correlations between predicted and observed performance in the Olympics clearly favoured race placing (0.78) over race time (0.65) and percent behind the winner (0.63) for women, whereas race placing was clearly inferior (0.14) to percent behind the winner (0.30) and race time (0.46) for men. All three methods are potentially useful and need further investigation in skeleton and other sports.