Jeremy Hellickson and Re-Defining BABIP

There aren’t many mainstream writers out there who are willing to tackle sabermetric concepts with players, so I have to give Marc Topkin from the the Tampa Bay Times a tip of the hat for mentioning BABIP to Jeremy Hellickson. Not only did he mention a sabermetric statistic to a player, but he brought one up that makes Hellickson look bad:

“Yea, I just got lucky on the mound,” Jeremy Hellickson says dryly. “A lot of lucky outs.” […]

“I hear it; it’s funny,” Hellickson said, not quite sure of the acronym. “I thought that’s what we’re supposed to do, let them put it in play and get outs. So I don’t really understand that. When you have a great defense, why not let them do their job? I’m not really a strikeout pitcher; I just get weak contact and let our defense play.”

First of all, I have to agree with Craig Calcaterra on this one: I couldn’t give a rats patootie if Hellickson knows about or understands BABIP. Sabermetrics is a field most helpful to front office personnel and managers, and while some players find it useful, players don’t need be saberists in order to be good players. And anyway, it’s never going to be the successful players that stumble upon sabermetrics; it’s always going to be the borderline players, the ones looking for any sort of possible advantage to help them get ahead. So should I be annoyed that Hellickson is poo-pooing BABIP? No, not in the least. Good on him.

Instead, this article caught my eye for a different reason: it refers to BABIP primarily as a measure of luck. Hellickson had a low BABIP, which therefore meant he was lucky on balls in play last year. Any pitcher with a low BABIP is therefore “lucky”, and any pitcher with a high BABIP is “unlucky”.

This is a common perception about BABIP, and one that used to be in favor among sabermetric circles. Heck, I subscribed to this philosophy three or four years ago, and I used “luck” as a quick way of describing BABIP to the uninitiated. But these days, that’s an outdated mindset and, quite frankly, misleading.

BABIP is one of the most important sabermetric concepts, but it’s also one of the most misunderstood. What does BABIP tell us? What doesn’t it tell us? Let’s explore.

In the FanGraphs Library, I state that there four main variables that affect a player’s BABIP: defense, talent level, skill set, and luck.  Obviously, pitchers who play in front of good defenses are going to have more of their balls in play turned into outs, and that’s especially true if managers know how to position their defenders optimally (see: Maddon, Joe). That’s an easy variable to see, and it’s why the Rays had an MLB-best .265 team BABIP last season. Not a single starting pitcher on the Rays last season had a BABIP higher than .284, considerably lower than the MLB average rate of .293.

Next, you have to consider how a player’s talent level affects their BABIP. If I were to step into a major league game, there’s no way I’d post a .300 BABIP; players would rip the ball off me, and I’d be lucky to get an out. The only pitchers who will survive in the majors are those that are good enough to retire major league hitters at a league-average rate. Pitchers may see their talent levels decrease if they’re playing through an injury, and their BABIPs will typically spike as a result.

Different skill sets can also influence BABIP rates. Pitchers with high strikeout rates tend to generate weaker contact and, therefore, allow fewer hits on balls in play. The same is generally true of relievers, as they can dial up the intensity over shorter outings. Also, fly balls fall in for hits less often than ground balls, and extreme ground ball pitchers (55% and above) are better at inducing ground balls that are easy for their defense to turn into outs.

And then, after all these other variables, comes luck. There are always going to be times when a pitcher performs well, but weakly hit balls sneak through the infield and multiple bloop hits fall in back to back. That happens; it’s part of the game. Over the course of a full season, some players may find themselves on Lady Luck’s bad (or good) side more often than others, and that’s something we simply can’t quantify. In my opinion, though, I think that luck has a lot less to do with BABIP fluctuations than we tend to assume.

So instead of the word “luck”, I prefer to use “random variation” to describe BABIP. If a pitcher posts an abnormal BABIP, it doesn’t necessarily matter why that player’s BABIP was so extreme — it’s still liable to regress. Based on all the above criteria, certain pitchers will regress toward different means. Strikeout pitchers will do slightly better than non-strikeout pitchers. Players pitching in front of good defenses should have lower BABIPs than players on defensively-challenged teams. These variables give each pitcher a different “expected” BABIP, with most major-league starters falling within the .275-.300 range.

If you look closely at Jeremy Hellickson, his .223 BABIP from last year doesn’t seem quite that scary going forward. He’s playing in front of a spectacular defensive team, so we should already expect his BABIP to be closer to .270-.280 than .290. Not only that, but Hellickson is an extreme fly ball pitcher, and fly balls fall in for hits less often than grounders. He also has shown a tendency in his young career to generate an extreme number of infield pop-ups, a la Jered Weaver (.276 career BABIP). While Hellickson may not be a high strikeout pitcher, these variables suggest we should expect a BABIP closer to .270 than .290 — a third less regression than he’d see if he regressed all the way to league-average.

It may look simple at first blush, but BABIP is actually one of the more complex sabermetric statistics. It’s not nearly as simple or cut-and-dried as many make it out to be, and I wouldn’t be surprised if we decrease the importance of “luck” even further once HITF/x data becomes available (if it ever does). So the next time you see or hear someone refer to BABIP as a luck statistic, be sure to mention that luck has little to do with it — it’s random variation.