Running the Numbers: Modeling Stolen-Base Decisions in Major League Baseball After the 2023 Rule Changes

Abstract

The 2023 MLB rule changes, namely larger bases and limits on pitcher disengagements, substantially altered the strategic landscape for stolen bases. This thesis examines how these changes affected both the decision to attempt a steal and the likelihood of success by developing a modeling framework for stolen-base behavior. Using event-level play-by-play data from the 2021–2025 Major League Baseball seasons combined with player-level attributes from Statcast, I estimate hierarchical logistic regression models that first estimate the probability of a steal attempt conditional on a steal opportunity and then estimate the probability of success conditional on an attempt. These models incorporate situational factors such as count, outs, inning, score differential, and disengagements, as well as player characteristics, including sprint speed and average pitch flight time, and account for persistent player-level tendencies through random effects. Comparing pre- and post-rule-change periods, the results show that the rule changes substantially increased runners’ willingness to attempt steals, while producing only modest improvements in success. This suggests that the primary impact of the new rules has been to shift strategic decision-making rather than fundamentally altering execution difficulty. The analysis also reveals significant heterogeneity across players, particularly in attempt behavior: a large share of variation in attempt rates remains even after controlling for speed and game context, while variation in success rates remains comparatively small. These findings indicate that differences in stolen-base outcomes are driven more by strategic decisions about when to run than by differences in execution ability. Player-level analyses further highlight that successful baserunning can arise through multiple strategies, ranging from broad aggressiveness to highly selective decision-making, and suggest that detailed situational awareness beyond what can be captured by the models may play an important role in determining success. Overall, this study provides a data-driven framework for understanding stolen-base decision-making in the modern Major League Baseball environment and offers insight into how the 2023 rule changes have reshaped strategic behavior on the basepaths.