BAYESIAN ANALYSIS OF STRIKE LIKELIHOOD IN BASEBALL: EVALUATING THE IMPACT OF PITCH DYNAMICS AND BATTER CHARACTERISTICS

Abstract

In baseball, a pitch resulting in strikes is very important for pitchers as it shapes count leverage, controls games, and also dictates overall command. To maximize these advantages, pitchers must understand how pitch characteristics and context influence the likelihood of a strike outcome. This study examines those factors by focusing on pitch selection, movement, and matchup dynamics. Using a Bayesian logistic regression model, a stratified sample of 4,593 pitches from over 1.4 million MLB records was analyzed to determine how pitch characteristics and batter-pitcher handedness influence the probability of a pitch resulting in a strike. The findings show that vertical location (plate_z) is the most important factor with pitches higher in the zone, more likely to result in strikes. Sliders and Sweepers were particularly effective at generating strikes, especially in full counts or two-strike situations, and their effectiveness varied with batter side-of-the-plate matchups. While batter side of the plate does influence strike likelihood, its impact depends heavily on pitch type, count pressure, and other contextual variables. On the other hand, pitches at the bottom edge of the zone (sz_bot) had a lower chance of resulting in strikes. By using Bayesian posterior distributions and contrast analysis, this study goes beyond simple correlations by providing a probabilistic, interpretable framework to understand pitch outcomes. These insights provide real-world applications for pitch design, scouting, and in-game strategy. Unlike previous research, this work incorporates strike zone variables, pitch movement, and handedness matchups through a causal lens in providing a richer and more actionable understanding of what really drives pitches to result in strikes.