2026-05-21T04:13:54Zhttps://keep.lib.asu.edu/oai/requestoai:keep.lib.asu.edu:node-2006642025-06-05T21:57:53Zoai_pmh:alloai_pmh:repo_items200664
https://hdl.handle.net/2286/R.2.N.200664
http://rightsstatements.org/vocab/InC/1.0/
http://creativecommons.org/licenses/by-nc-sa/4.0
2025-05
42 pages
Thompson, Nicholas
McIntosh, Daniel
Eaton, John
Barrett, The Honors College
School of Mathematical and Statistical Sciences
This Honors Thesis utilized a dataset of various by-month statistics for MLB players, compiled by Baseball Prospectus writer Robert Orr, to develop a classification model for predicting whether or not a player’s On-Base Plus Slugging (OPS) will increase or decrease across 2-month intervals. Model features include 0.05-level statistically significant changes in several statistics across 2-month intervals. Features were selected using background information relevant to baseball and MLB, simple linear regression to determine the stickiness of each statistic of interest between 2-month intervals, T-Tests to determine OPS predictive viability, and conditional independence tests to eliminate redundant features.
The best logistic regression model for 2-month OPS changes achieved 69% accuracy and 72% concordance in predicting the sign of change in OPS for hitters who observed evidence of a skill change. The best logistic regression model for 6-month OPS changes achieved 71% accuracy and 79% concordance in predicting the sign of change in OPS for hitters who observed evidence of a skill change, while also considering their OPS in the prior 6 months.
Further analysis showed preliminary evidence of player subtypes and other statistics evidential of skill changes that could be used for improving the model in the future. Evidence of skills that translate from the minor leagues to the major leagues also motivates future work and applications.
Baseball
Baseball Analytics
MLB
Hitting
Machine learning
Classification
Regression
The Latest and Greatest: Determining Sustainable MLB Hitter Outcomes from Sudden Skill Changes