In baseball, heat maps, which visualize a batter`s ability across regions in and around the strike zone, play an important role in baseball commentary and scouting reports. We represent the stochastic process underlying these heat maps as a spatial Gaussian field with isotropic covariance. Spatial interpolation (kriging) relies on the assumption of a known covariance function, but in reality the parameters of the covariance are typically estimated from the data. Our simulation study, based on a known covariance function, helps to understand and explain the spatial dependence of the process and allows us to produce improved heat maps.
© Copyright 2015 Journal of Quantitative Analysis in Sports. de Gruyter. All rights reserved.