BPro: Silver: Evaluating Pitcher Hitting

Swayed by Owings? 

Don’t be a fool
Remember the Buhl!

In the six years that I’ve generated PECOTA forecasts, I’ve never bothered to run hitting projections for pitchers. In fact, I’ve regarded pitcher hitting as something of a nuisance; I specifically screen out any pitchers so that they won’t be selected as comparable players. This isn’t an aesthetic judgment by any means—watching pitchers try (and fail) to hit is one of my favorite pastimes. But since even the pitchers who make 35 starts a year won’t usually get more than 80 or 90 plate appearances, I’ve generally figured that it wasn’t quite worth the trouble.

Still, if you aren’t accounting for pitcher hitting, you’re unambiguously cheating some guys out of their value. For example, Micah Owings had a VORP of 16.5 last season, so his offense did more to contribute to his team’s success than actual, real-life hitters like J.D. Drew, Ivan Rodriguez, and Rafael Furcal. But how much of that performance was actual skill rather than a fluke of small sample size? Pitcher hitting suffers from the monkeys-typing-Shakespeare problem: take enough pitchers, give them all 70 at-bats or so, and two or three of them are going to wind up hitting .320 with a couple of long balls, perhaps simply by a stroke of luck.

To address this question, I developed a regression-based system for projecting hitter performance that I’ll call SPHPS (Simple Pitcher Hitting Projection System), which isn’t a catchy anagram at all, but has the virtue of being a palindrome. This system is derived solely from the hitting statistics of pitchers; you don’t want to include position players in the same dataset because then you’ll regress toward the mean established by position players, when pitchers and position players are selected to play in the major leagues based on totally different skill sets. (The gap between pitcher and position-player hitting has grown steadily since the dawn of baseball time).