The success of critical velocity protocol on predicting 10000 meters running performance

Background and Study Aim The study aims to evaluate which of the critical velocity (CV) estimates of the three widely used models and the best-fit model successfully predict the running performance of 10000 meters. Material and Methods The group of participants in this study consisted of 11 British endurance athletes. The CV estimations were obtained from the models with the athletes' running velocity and exhaustion times of 1500, 3000, and 5000 meters (m). The information was taken from a website where the results of the British athletes are recorded. In terms of selecting endurance athletes, the data of the athletes who ran 1500 m, 3000 m, 5000 m, and 10000 m in the same two years were included in this study. By fitting the data into mathematical models, the CV estimates of the three mathematical models and the individual best-fit model were compared with the 10000 m running velocity. The CV estimates were obtained by fitting the relevant data on the running velocity, exhaustion time, and running distance of the three running distances of athletes to each of the three mathematical models. Results 10000 m running velocity and times of the athletes corresponded to 19.65 +/- 1.26 km/h and 30.4 +/- 1.94 minutes, respectively. The CV values obtained from the three mathematical models and 10000 m running velocity were similar (p > 0.05). Although the lowest total standard error levels were obtained with the best individual fit method, the 10000 m running velocity was overestimated (p < 0.05). Conclusions Three mathematical models predicted 10000 meters of race velocity when an exhaustion interval between 2-15 minutes was used. Even though the mathematically most valid CV value was obtained with the best individual fit method, it overestimated the 10000 m running velocity. When comparing the values of CV and the velocity of running 10,000 meters, our study suggests using the model. This is because the model has the smallest effect size, and there is no statistically significant difference in the total standard error level between the model and the best-fit model.