An Analysis of Pacing Profiles in Sprint Kayak Racing Using Functional Principal Components and Hidden Markov Models

TL;DR

This paper uses functional principal component analysis and Hidden Markov Models to analyze and categorize pacing profiles in sprint kayak racing, revealing how athlete pacing evolves over their career.

Contribution

It introduces a novel combination of FPCA and HMM to classify and track pacing profile changes in athletes over time.

Findings

Pacing profiles can be effectively categorized into four types.

Athletes show significant changes in pacing as they mature.

Higher dropoff in pacing observed in development athletes.

Abstract

This study analysed sprint kayak pacing profiles in order to categorise and compare an athlete's race profile throughout their career. We used functional principal component analysis of normalised velocity data for 500m and 1000m races to quantify pacing. The first four principal components explained 90.77% of the variation over 500m and 78.80% over 1000m. These principal components were then associated with unique pacing characteristics with the first component defined as a dropoff in velocity and the second component defined as a kick. We then applied a Hidden Markov model to categorise each profile over an athlete's career, using the PC scores, into different types of race profiles. This model included age and event type and we identified a trend for a higher dropoff in development pathway athletes. Using the four different race profile types, four athletes had all their race profiles throughout their careers analysed. It was identified that an athlete's pacing profile can and does change throughout their career as an athlete matures. This information provides coaches, practitioners and athletes with expectations as to how pacing profiles can be expected to change across the course of an athlete's career.