Kangyi (Ken) Peng

Title: Bayesian Approaches for Critical Velocity Models
Date: November 17, 2021
Time: 4:00 pm (PDT)
Location: Remote delivery

Abstract

In sports science, critical power and related critical velocity models have been widely inves-tigated, and are being increasingly applied to field-based team sports. A challenge associ-ated with these models is that laboratory experiments which yield accurate measurements of maximal sustainable velocity are expensive. Alternatively, inexpensive field data (from training and matches) are being used to fit such models. However, with field data, the de-pendent variable concerning maximum sustainable velocity is reliably calibrated only for short time durations. This paper develops methods where field data based on short time du-rations is combined with prior knowledge to fit the three-parameter critical velocity model. This is accomplished in a Bayesian framework where Markov chain methods are required for model fitting and inference.

Keywords: correlated observations; informative priors; wearable devices; critical velocity