Race course characteristics are the most important predictors in 48 h ultramarathon running

Abstract

Abstract Ultra-marathon running - where races are held in distance-limited (50 km, 50 miles, 100 km, 100 miles, etc.), time-limited (6 h, 12 h, 24 h, 48 h, 72 h, etc.), and multi-stage races - is gaining in popularity. However, we have no knowledge of where the fastest 48-hour runners originate and where the fastest 48-hour races are held. This study tried to determine the origin of the fastest 48-hour runners and the predictor factors associated with 48-hour ultra-marathon performance, such as age, gender, event country, country of origin and race course specific characteristics. A machine learning (ML) model based on the XG Boost algorithm was built to predict running speed from the athlete´s age, gender, country of origin, where the race occurs and race course characteristic such as elevation (flat or hilly) and surface (asphalt, cement, granite, grass, gravel, sand, track, or trail). Model explainability tools were then used to investigate how each independent variable would influence the predicted result. A sample of 16,233 race records from 7,075 unique runners originating from 60 different countries and participating in races held in 36 different countries between 1980 and 2022 was analyzed. Participation was spread across many countries, with USA, France, Germany, and Australia at the top of the participants’ rankings. Athletes from Japan, Israel, and Iceland achieved the fastest average running speed. The fastest races were held in Japan, France, Great Britain, Netherlands, and Egypt. The XG Boost model showed that elevation of the course (flat course) and the running surface (track) were the variables that had a larger influence on the running speed. The country of origin of the athlete and the country where the event was hold were the most important features by the SHAP analysis, yielding the broader range of model outputs. Men were ~ 0.5 km/h faster than women. Most finishers were 45–49 years old, and runners in this age group achieved the fastest running speeds. In summary, elevation of the course (flat course) and the running surface (track) were the most important variables for fast 48-hour races, whilst the country of origin of the athlete and the country where the event was hold would lead to the broadest difference in the predicted running speed range. Athletes from Japan, Israel, and Iceland achieved the fastest average running speed. The fastest races were held in Japan, France, Great Britain, Netherlands, and Egypt. Any athlete intending to achieve a personal best performance in this race format can benefit from these findings by selecting the most appropriate race course.

Keywords

• Ultra-endurance
• Performance analysis
• Nationality
• Machine learning