The aim of the present study was to predict triathlon performance at Paris 2024 Olympic Games using conventional statistics and a machine learning-based approach. It was hypothesized that both predictive models would grant sufficiently accurate results. Data were extracted from the API service on the World Triathlon website. A custom Python code was written for the analyses during data collection. Conventional statistics and machine learning analyses were performed by creating a Jupyter Notebook via Google Colab. Data for machine learning were divided into training (80%), and testing (20%). Run time was the best-predicted discipline for males (average difference: -0.21% ± 5.45%), and total time was the best-predicted variable for females (average difference: -5.43% ± 3.81%). For males, the linear regression based on Olympic Games races data was the most accurate technique overall (average relative difference: -3.03% ± 10.79%). For females, TensorFlow achieved the best precision (average relative difference: -1.95% ± 10.43%). Both techniques are reliable approaches for predicting performance in Olympic triathlon. Based on our findings, statistical methods can be effectively employed by coaches to predict individual discipline performance and overall race times. When comparing statistical techniques that considered all data, machine-learning techniques showed better predictions than conventional statistics.
© Copyright 2026 Journal of Human Sport & Exercise. University of Alicante. Alle Rechte vorbehalten.