Reproduce “The Impacts of Increasingly Complex Matchup Models on Baseball Win Probability” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Tristan Mott, Caleb Bradshaw, David Grimsman, Christopher Archibald
Researchers from Brigham Young University developed and evaluated a series of hierarchical Bayesian models to predict baseball plate appearance outcomes, finding that more accurate models can translate to approximately one additional win per 162-game season through improved in-game strategic decisions. Their most advanced model, which includes pitcher, batter, recency, and player-specific base running tendencies, also generated win probability estimates consistent with market betting lines, showing potential for positive returns with a confidence threshold.
Key insights
- More accurate hierarchical Bayesian matchup models, particularly those incorporating individual player tendencies and recent performance, lead to tangible gains in win probability through improved strategic decision-making.
- The impact of model features on win probability does not always directly correlate with their effect on log loss; for example, recency significantly influenced strategy while player-specific base running had less discernible impact on simulated strategic choices.
- High-fidelity matchup models can produce win probability estimates that align with external market expectations and can potentially identify betting value with appropriate confidence thresholds.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.