Optimal Learning for Optimal Rowing: Maximizing Mechanical Efficiency

Abstract

The goal of this paper is to apply sequential decision analytics to learn optimal rowing stroke mechanics. We discuss the techniques used to help quantify the rowing stroke and varying body mechanics that different rowers employ. A sequential decision mathematical model is developed using five main components: states variables, decision variables, exogenous information, transition functions, and the objective function. The backbone of this model is the concept of correlated beliefs and prior belief distributions. Using an online learning algorithm, we simulate many races under this mathematical model. In the process of simulating many races, we compute and compare the performance of different policies to learn the best way to make a decision under this model formulation. Once we find an optimal policy that makes the best decisions, we run more simulations of races to find an optimal technical strategy that results in the fastest expected race times. Using sequential decision analytics, we seek to objectify the most subjective components of the rowing stroke.