Pitching by the Numbers: Debasing BABIP

If there’s one thing I’d like all my readers to do, it’s stop focusing on BABIP. It’s not a very meaningful metric in our game, as I will happily illustrate again by focusing on the much more meaningful isolated slugging allowed (ISO, slugging average minus batting average). In other words, it's not the hit but the hit quality that matters.

Think about it: BABIP focuses mostly on singles, the least important hits. Those are the ones that have the highest BABIP variance, meaning that they can be chalked up more to luck than skill. But singles don’t really hurt us (by hurting our pitchers) that badly. The bigger hits do and BABIP doesn’t even factor in homers.

The pushback I’ve gotten on this has been fierce since I introduced the idea of using it in 2012 to explain why Johnny Cueto is so good at preventing runs regardless of how his BABIP varied. One of the arguments is that this is some sort of tautology. Yet the preseason focus with Cueto that year was that his BABIP would surely correct so forget that great 2011 ERA. Then it did but his ERA was still great and that’s where I came in.

The other argument is from the hard-core stats guys who say that ISO takes so long to normalize and is thus impossible to predict. In other words, it merely describes good pitching but what comes tomorrow for pitchers who haven’t had YEARS to prove that they are great at ISO, who knows. And even if they have proven it, their ISO can still vary widely.

I’m not writing a white paper here. Whether ISO correlation meets some quant standard is not important to me. It’s binary: is a pitcher good at it or bad at it? In other words, can we expect him to be above or below average — catch the ISO tailwind with ERA or have to pitch into its headwind?

One of the arguments I counter with is that if ISO takes 600 innings to become reliable (or whatever the number is purported to be) than why are relievers predictably good or bad at it every year with only a fraction of those innings? Mariano Rivera’s excellent ISO was bettable, for example. How excellent Rivera’s ISO would be may have varied significantly, but it was never going to be bad.

Let’s look at this year’s leaders in ISO allowed before continuing this defense:

Focusing on the top 10, the super elite in ISO, I note that, prior to the season, based only on how they pitched recently and not even their entire careers, I would have predicted eight of those 10 to be good at it. How many stats can you say that about? It sure seems like an argument against randomness/unpredictable variance to me. Archer and Quintana weren’t bad at it but weren’t good, either. But I do note that, like Cueto, Archer’s BABIP corrected mightily as many predicted but without the corresponding ERA impact because of ISO. Even if it’s tautology, can we at least explain correctly why this is happening instead of just ignoring it and moving on the same, lame BABIP argument next time?

I don’t want to trouble you with another chart, but it’s even more predictable at the bottom of the ISO rankings. There, it would have been very easy using smaller, recent samples to predict poor ISO performance this year from EVERY ONE of the 10 of the worst ISO pitchers (in order, Marco Estrada, John Danks, Dan Haren, Jake Peavy, Eric Stults, Chris Young, Edwin Jackson, Wei-Yin Chen, Tim Lincecum and R.A. Dickey). So prior to the season, I would have predicted that 18 of the 20 best and worst ISO pitchers would be good or bad at it. That’s a tool in the toolbox, kids.

And look how badly FIP ERA and xFIP continue to let us down with those bottom-dwelling ISO pitchers. We’re trying to be smart, but it’s making us stupid by normalizing the hits equally, even the bigger hits that we’d all agree are less due to bad luck than hard contact. If we can’t fault pitchers for hard contact, what are we doing here? So people will buy into these pitchers for other reasons, hoping their ERAs will correct in accordance with the FIPs but it does not happen and it will not happen because of ISO.

The other criticism is that ISO is some substitute for simple fly ball and ground ball rates. Yes, the ground ball pitchers are going to be better at ISO generally than the fly ball pitchers. But if that’s so obvious, why are we still focusing on BABIP and the FIPs (also my favorite 1970s soul group) in predicting them? Why confuse us all with those stats and completely ignore ISO?