Original title: Auditing Contextual Bias in Human Ball-Strike Calls Using KBO's Automated Umpiring Transition
Kichang Lee, JeongGil Ko · September 2026
Baseball umpires must make quick calls, but human judgment can bend under pressure. Researchers studied whether human umpires alter their strike zones depending on the situation. They analyzed pitches in the Korean Baseball Organization from 2021 through part of the 2026 season. They compared human calls to calls made by an Automated Ball-Strike system, a computer setup that tracks pitches robotically. The researchers looked at edge-of-the-zone pitches that batters did not swing at. They found clear bias in game counts. With two strikes, human umpires were far less likely to call a strike. With three balls, they called more strikes. Humans also called more strikes late in games, especially in tight contests, while the automated system stayed flat. Other suspected biases were weaker. Player salaries showed only suggestive results, while home-team advantages were mostly absent. Overall, human biases appeared strongly in specific moments rather than everywhere.
In a cooking competition, judges might give higher scores to dishes that are similar to previous winning dishes, rather than judging each on its own merit. This happens because past results influence current decisions, even unintentionally. In this paper, human umpires show similar behavior by calling more strikes in certain game counts, while the automated system judges each pitch independently.
This matters because baseball games can hinge on close calls, and if umpires are influenced by the game state, it affects fairness for both pitchers and batters. Knowing where biases exist helps decide whether automated systems can improve officiating.
The study is limited to KBO data and may not generalize to other leagues. It only shows biases in specific contexts like count pressure and game progression, not all potential biases. Some evidence, like salary-based reputation, is suggestive but depends on proxies and may not be definitive.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Auditing Contextual Bias in Human Ball-Strike Calls Using KBO's Automated Umpiring Transition — Kichang Lee, JeongGil Ko
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.