Modeling the probability of a batter/pitcher matchup event

Abstract

We develop a hierarchical Bayesian model for predicting whether a specific batter reaches on-base from a matchup with a specific pitcher. Using less than 100 lines of historical matchup data from a particular batter/pitcher matchup, our proposed model allows estimating the probability of the event occurring in the matchup of corresponding players. We also show that the defense ability of pitcher's team is negatively correlated to the probability that the opponent team's batter reaches on-base and extend our model with a new variable. The proposed model is estimated from Markov chain Monte Carlo method. We compare the performance of our model to the log5 model and generalized log5 model which are commonly used methods for modeling the probability of the event in baseball.