A machine learning approach to triathlon component analysis

This paper presents work on using machine learning approaches for analyzing triathlon components for talent identification and predicting performance patterns of medalists. Triathlon is a high profile multi-component sport which requires strategic understanding and planning with respect to each of the three sports: swimming, cycling, and running, to achieve the best overall standing. We carried out machine learning decision tree-based analysis and probabilistic modeling on triathlon historical data since 1989 to investigate what minimum time gap behind the lead athlete (for each event) still results in a high likelihood of medaling. We found the minimum time differentials from the leader in each event, as well as the overall differential time that still results in a large chance of medal winning.