MLB Automatic Runner Selection Reproducibility Package

This reproducibility package accompanies the manuscript “Context-dependent bunt-attempt selection under Major League Baseball's automatic-runner rule.”

The study examines associations between actual batting role, score state, and recorded bunt-attempt selection in 2020–2025 MLB regular-season automatic-runner opportunities. The archive includes derived data, analysis code, archived model outputs, a frozen analysis manifest, dependency information, reproduction instructions, and SHA-256 checksums.

The automatic-runner sample contains 3,220 eligible opportunities. The primary logistic model uses 2,772 opportunities in three batting-role and score-state categories. Outcome, ability, and temporal-policy analyses are secondary or exploratory. The results describe observed action selection and do not establish causal tactical effects.

Version 1.0.1 corrects the archive directory structure and adds existing reproducibility files. These packaging changes do not alter the frozen analysis or reported estimates. Original Retrosheet event files and third-party executables are not included.

The associated manuscript is intended for submission to the Journal of Sports Analytics; this record does not indicate journal acceptance or publication.