A hitter is 5-for-10 against the guy on the mound, and the batter vs pitcher stats look like a green light. Hold on. Ten at-bats swing wildly on a few balls in play. You will see shrinkage in action: start with a Beta prior centered at .250, add five hits in ten at-bats, and the estimate lands near .278 instead of .500. Then comes the harder question of when matchup history adds real signal, such as a high-spin breaking ball against a particular hitter, and how to test that against handedness, park, location and changing skill. The workflow builds from a population baseline upward, so you only trust head-to-head history after it proves it earns a place.
Learning objectives
Interpret the technique, reproduce the example, describe limitations, and apply it to a simulated research workflow.
