Model-based estimation of baseball batting metrics

Area of Science:

• Sports Analytics
• Statistical Modeling
• Baseball Performance Metrics

Background:

• Traditional baseball batting metrics often analyze players in isolation.
• Improving the accuracy of player statistics is crucial for performance evaluation and comparison.
• Information sharing across players can enhance statistical inference.

Purpose of the Study:

• To introduce and evaluate a weighted likelihood approach for modeling baseball batting outcomes.
• To compare this new approach with a semi-parametric Bayesian method using the Dirichlet process.
• To estimate commonly used baseball batting metrics with improved accuracy.

Main Methods:

• Utilized a weighted likelihood approach with Minimum Averaged Mean Squared Error (MAMSE) weights.
• Implemented a semi-parametric Bayesian approach based on the Dirichlet process for comparison.
• Applied methodologies to 2018 Major League Baseball (MLB) batters data.

Main Results:

• The weighted likelihood approach allows for effective information sharing among players.
• This sharing of data leads to improved statistical inference for individual batters.
• Both introduced methods were demonstrated and compared using real-world MLB data.

Conclusions:

• The weighted likelihood approach offers a robust method for modeling baseball batting performance.
• Information sharing across players significantly enhances the estimation of batting metrics.
• The study provides a valuable framework for advanced sports analytics and statistical modeling in baseball.