DIBS and the future of batting statistics

PHOENIX — I’m hitting a few ballparks this week, but last week I attended the annual SABR Analytics Conference, which has become the place if you’€™re of that particular bent. Just for an idea: I always keep a small notebook handy, so I don’™t forget things I might want to write about someday. During the three days of the conference, during which I attended a number of talks, met with some old friends and made some new ones, I filled eight pages. Which is a personal record.

I’m not going to recap the entire conference; you can get caught up some here. I would like to write a few short pieces this week about things that particularly intrigued or inspired me. Starting with Ben Jedlovec’€™s presentation about Defense Independent Batting Statistics (DIBS).

We’€™ve all been cognizant of DIPS for … gosh, it’s been nearly 15 years now. But for whatever reasons, we don’€™t usually pay nearly as much attention to DIBS. Maybe because there’s not a single baseline for DIBS, as there is for DIPS.

What Jedlovec has done, though, is drill about as deeply into hitters’ performances as possible, and come up with something that’s more useful, more predictive, than even the more sophisticated numbers we usually reference.

What are we talking about? Let’€™s say you hit a line drive into right-center field, maybe 378 feet from the plate. A real screamer. That’s usually a double or a triple, right? In fact, based on the location and trajectory of the batted ball, we can estimate that your batted ball is usually worth 0.63 runs (that’€™s a made-up number, by the way).

Except this time Lorenzo Cain made a tremendous play. You’re out and nobody on the bases advanced, so in this case your efforts were worth exactly 0.00 runs.

Which doesn’€™t seem fair, does it? Well, nobody cares. For our purposes, what matters is that your 0.00 runs isn’t at all predictive. That big fat zero is actually quite deceptive, because it tells us nothing about your fundamental performance.

Well, DIBS does it the other way; DIBS gives you 0.63 runs for your screamer.

Now, it’€™s true that DIBS is hardly a new idea. Years and years ago, Paul DePodesta talked about stastistical derivatives; that is, looking not at outcomes like hits and walks and the like, but rather the little pieces and probabilities that contributed to them. What’€™s different now is that batted-ball data is more precise than it was 15 years ago when I started writing about this stuff. Smart people have been looking at the derivatives for at least that long. What’€™s different is the data. And the teams — or the teams that give a damn, anyway — have even better data than Jedlovec.

The rest of us can come pretty close, though. I mean, we already had the raw statistics, which are pretty good. Then we had BABiP, which can be useful for hitters as well as pitchers, along with more granular batted-ball percentages. But somebody really should be publishing DIBS, whether Jedlovec’s version or someone else’s. I’m surprised that it hasn’t happened yet.

Will these numbers have an appreciable impact on the accuracy of projections, though? I’€™m not sure. Studying the question a few months ago, Rob Arthur demonstrated that most projection systems come up with the same answers, and that a high percentage of players are relatively easy to project (absent debilitating injuries, anyhow). But there’€™s a significant percentage of players every season whom the projections completely miss on. And it’s not at all obvious that we’ll ever knock that percentage down much. Because (as I should hasten to add) at least some of the systems already consider derivatives. Is Jedlovec’s method more sophisticated? Maybe. I’d like to see it applied to BIS’€™s own projections, and we’ll see if they do appreciably better than the others.

Ben did offer one great example, by the way. For four or five years straight, David Ortiz’€™s OPS lined up almost perfectly with his DIBS OPS. And of course he was an outstanding hitter from 2011 through ‘€™13. In 2014, he was somewhat less outstanding. However, according to DIBS he was almost exactly the same hitter last season — yes, despite being 38 years old — but just hit in pretty tough luck. Which obviously bodes well for his 2015. Even at 39. So we’€™ll see about that, although of course one player in one season can’€™t tell us much of anything at all.

Some of you might already be wondering: "Hey, what about StatCast?"

Well, yeah. When the player-tracking system is fully deployed, teams (at least) will really be able to nail the derivatives, because of course then they’€™ll know everything about that batted ball and the likelihood of various subsequent outcomes.

It’s not at all clear how much of that data will be publicly available. But the smart guys can get pretty close with what’s available right now. And I’ll be surprised if Ben’s or some other version of DIBS isn’€™t showing up on your favorite statistics page by next March.