Analysis of over 1 million race records shows runners from East African countries as the fastest in 50-km ultra-marathons

Abstract

The 50-km ultra-marathon is a popular race distance, slightly longer than the classic marathon distance. However, little is known about the country of affiliation and age of the fastest 50-km ultra-marathon runners and where the fastest races are typically held. Therefore, this study aimed to investigate a large dataset of race records for the 50-km distance race to identify the country of affiliation and the age of the fastest runners as well as the locations of the fastest races. A total of 1,398,845 50-km race records (men, n = 1,026,546; women, n = 372,299) were analyzed using both descriptive statistics and advanced regression techniques. This study revealed significant trends in the performance of 50-km ultra-marathoners. The fastest 50-km runners came from African countries, while the fastest races were found to occur in Europe and the Middle East. Runners from Ethiopia, Lesotho, Malawi, and Kenya were the fastest in this race distance. The fastest 50-km racecourses, providing ideal conditions for faster race times, are in Europe (Luxembourg, Belarus, and Lithuania) and the Middle East (Qatar and Jordan). Surprisingly, the fastest ultra-marathoners in the 50-km distance were found to fall into the age group of 20-24 years, challenging the conventional belief that peak ultra-marathon performance comes in older age groups. These findings contribute to a better understanding of the performance models in 50-km ultra-marathons and can serve as valuable insights for runners, coaches, and race organizers in optimizing training strategies and racecourse selection.

Figure 1

The number of runners and men-to-women ratio over the years.

Figure 2

Model interpretability charts for feature: Gender. (a) Predictions of race speed (km/h); (b) Average race speed (km/h); (c) Records count.

Figure 3

Model interpretability charts for feature: Age group. (a) Predictions of race speed (km/h); (b) Average race speed (km/h); (c) Records count.

Figure 4

Model interpretability charts for feature: Athlete country. (a) Predictions of race speed (km/h); (b) Average race speed (km/h); (c) Records count.

Figure 5

Model interpretability charts for feature: Event country. (a) Predictions of race speed (km/h); (b) Average race speed (km/h); (c) Records count.

Figure 6

Evaluation metrics and features importance. Sample size 1,398,845. XGBoost trees 500. MAE (km/h 1.39). Feature importance Event_country_ID 0.66, Athlete_gender_ID 0.23, Age_group_ID 0.07, Athlete_country_ID 0.05.