An analysis of the 6-hour ultra-marathon race using a machine learning approach

Mabliny Thuany¹Katja Weiss²David Valero³Elias Villiger²Marilia Santos Andrade⁴Pantelis Theodoros Nikolaidis⁵Volker Scheer³De Lira, CA⁶Rodrigo Luiz Vancini⁷Ivan Cuk⁸Lorin Braschler⁹Thomas Rosemann¹⁰Beat Knechtle^10,2*

• ¹Universidade do Estado do Pará, Belem, Para, Brazil
• ²Institute of Primary Care, University of Zurich, Zurich, Switzerland, Zurich, Switzerland
• ³Ultra Sports Science Foundation, 69130 Pierre-Bénite, France, Pierre-Bénite, France
• ⁴Federal University of São Paulo, São Paulo, São Paulo, Brazil
• ⁵University of West Attica, Athens, Greece
• ⁶Universidade Federal de Goiás, Goiânia, Goiás, Brazil
• ⁷Federal University of Espirito Santo, Vitória, Espirito Santo, Brazil
• ⁸University of Belgrade, Belgrade, Serbia
• ⁹University of Bern, Bern, Bern, Switzerland
• ¹⁰University of Zurich, Zürich, Zürich, Switzerland

Background: Ultra-marathon running popularity is increasing, with the 6-hour run being the shortest time-limited ultra-marathon. Since very little is known regarding the country from which the fastest 6-hour runners originate, the fastest age group, and where the fastest 6-hour race courses are located, this study aims to close this gap. Methods: A machine learning model based on the XG Boost algorithm was built to predict running speed based on the athlete´s age, gender, country of origin, and the country where the race takes place. Model explainability tools were used to investigate how each independent variable would influence the predicted running speed. To assess the impact of individual performance against the other variables under study, a Mixed Effects Linear Model was also built. Results: A total of 117,882 race records from 51,018 unique runners from 65 countries participating in races held in 56 different countries were analyzed. Participation is spread across a wide range of countries, with a high correlation between the country of origin and the country of the event. Most runners originated from Germany, Italy, France, the USA, and Sweden, with Europe (Belgium, Russia, Spain, Poland, Romania, and Lithuania), being the fastest. Most athletes competed in Italy, Germany, France, the USA, and The Netherlands. The fastest average running speeds were also achieved in European countries (Russia, Belgium, Poland, Netherlands, Romania, Croatia, and Lithuania). Conclusions: For athletes competing in a 6-hour ultramarathon, gender was the most important predictor, followed by the origin of the athlete, the age, and the race location. The 6-hour running event seems to be dominated by European athletes regarding both participation and performance.