BABIP vs. zBABIP at the Halfway Point
Since Voros McCracken pondered the meaning of BABIP back in 2001, much of sabermetric research has had an increased focus on volatility. This is especially important when you run projections — as I do from time to time when the mood strikes me — since that volatility has a way of confounding your prognostications. While the evidence suggests that hitters have much more of an actual BABIP “ability” than pitchers do, it doesn’t mean that such an ability is on the same firm ground as, say, plate discipline or the ability to crush the ball into an alternate universe. Even still, outliers tell us a lot even if we expect some of those outliers to remain outliers to some degree.
As odd as it still seems, we’re essentially at the halfway point of the 2020 season for most teams. Weird BABIP numbers don’t magically just work themselves out in a normal 162-game schedule, so we would expect them to do so even less when the season is only 60 games. In a situation like this, estimates of what BABIP a player “should” have based on their advanced data will have more relevance to future seasons than the actual BABIPs do.
One feature built into ZiPS — and into the next iteration of the in-season model — are “z” stats, ZiPS’ attempt to make sense of volatile numbers. For stats like pitcher homers, zHR is far more predictive than the actual number of home runs allowed is (the most predictive model using just HR and zHR weighs the latter about nine times that of the former). zBABIP for hitters isn’t quite on the same level, with an r-squared of only 0.54 historically, but it does still add a lot of information about which players exceeding or falling short of typical BABIP numbers “deserve” to do so. Read the rest of this entry »