Last week, Baseball Prospectus introduced a new metric for pitchers: Deserved Run Average, or DRA. By zeroing in on some key component statistics while still baking in some context, we can get a read on how much of what the D-backs’ pitchers’ stats look like thus far is chance or bad luck, and how much is not far off the mark. Sure, past performance is no guarantee of future success or lack of success, but it’s still very helpful to get a new kind of read on what pitchers have done thus far. Hint: Jeremy Hellickson‘s 5.20 ERA is… pretty accurate.

We may quickly find DRA to be a highly useful tool. In these parts, we’ve used an ERA estimator called SIERA quite a lot — it breaks pitching down to some component parts, and builds a number for what a pitcher’s ERA should have been. As noted on FanGraphs, it also ends up faring pretty well as a projection system, just by filtering out some of the noise from past results. But as noted in the in-depth discussion at Baseball Prospectus, DRA doesn’t choose between what has thus far been an either/or question: are we explaining what happened (as with ERA, RA9), or just what should have happened, based on what the pitcher did in terms of what he can control (SIERA, FIP)?

The thing I find most compelling about DRA is the point made that we just understand more about a pitcher’s role in run scoring. Defensive-Independent Pitching theory has been pretty simple, the first big innovation being that pitchers affect outcomes a lot less than we thought twenty or more years ago. At FanGraphs, the FIP statistic is exceedingly simple, based on walk rate, strikeout rate, and home run rate. That’s it. We learned more, and we could incorporate more, like how ground ball percentage works, how it affects BABIP when a pitcher’s ground ball or fly ball rate is particularly high, and how allowing some walks might be less than half as harmful to runs scored as twice as many walks, because some is not so bad, but too many is too many. That’s all baked into SIERA.

But we do know more now. We know more about the strike zone and catcher framing and blocking, about platoon splits and how pitcher and batter tendencies interact, about how to weigh different elements of defensive contributions. In making DRA, the Baseball Prospectus team has included some new innovations on base stealing and steal attempts, and how pitchers affect those things and how they affect run scoring. We’ve been equipped for some time to get more granular about strength of competition, but now know how those should fit together. Why not use all of this? More accurate reads means more reads, more quickly.

As the BP guys put it, you can start in “reverse” with outcomes and strip things out, like Baseball-Reference, and end up attributing to the pitcher things we don’t necessarily understand; or you can start bare bones and forward, and attribute to the pitcher only things you know he’s responsible for (mostly). DRA helps bridge that gap; working forward, we’re able to add so much information now that we have this new statistic that apparently does a great job of explaining the past while also giving us insight about what’s likely to happen moving forward. Go read BP’s intro to DRA, if you can — totally worth the price of admission.

My favorite part, from explaining the final calculation and why they’ve used Multivariate Adaptive Regression Splines:

As discussed above, pitchers who are pitching particularly well or poorly have a cascading effect on other aspects of the game, including base-stealing. Moreover, there is a survival bias in baseball, as with most sports: pitchers who pitch more innings tend to be more talented, which means they end up being starters instead of relievers or spot fill-ins. The power of MARS is it not only allows us to connect data with hinged lines rather than straight ones, but that it allows those hinges to be built around the most significant interactions between the variables being considered, and only at the levels those interactions have value. MARS also uses a stepwise variable selection process to choose only the number of terms sufficient to account for the most variance.

Yes!! Putting things into boxes is much more helpful than not doing so. But we can make more complicated shapes; the extent to which walks affect pitchers’ outcomes is a really good example. Allow none, and you’re doing something wrong — so the very few shouldn’t matter in some kind of straight line with the many. These guys are the badasses of gory baseball math, and I’m not going to pretend I understand all of it. But step by step, it all makes sense. And it sure as hell sounds like a good idea.

Jeremy Hellickson might not agree. So far, Hellickson has a 5.20 ERA. His DRA : 5.22. It’s not bad luck, unless you count drawing the Giants twice along with the Dodgers, Pirates and Padres bad luck. I wouldn’t.

Hellickson has tried a somewhat different approach his last two times out: throwing the four-seam a lot more, close to 60% in both starts. For a guy throwing four pitches, that’s pretty high. For Hellickson, it’s really a question of four-seam or sinker, and I guess you don’t really get bonus points for being unpredictable on what you rely on — it’s probably more important to be unpredictable within at bats. I’m not a pitcher, though, and I’m not Hellickson. Another explanation is that on some days, he’s just not feeling one fastball or the other. I haven’t heard that with respect to fastballs — more about breaking balls and changeups — but I can’t say it’s not possible, either.

On a per-pitch rather than per-PA basis, line drive percentage suggests that Hellickson has got the balance pretty much right on this year. It’s exactly 5.97% for both the four-seam and curve, and 6.31% for the change, 6.82% for the sinker. Hitters are squaring up each of the four pitches at almost the same rate, and I think the default expectation there is that he’s de-emphasized the easier pitches to square just enough to make them unpredictable enough to match the better pitches. Optimized, people!

Only, the results are less than optimal. With Hellickson, we’ll always be counting on a good whiff rate on the change — but he’s had that, with hitters swinging and missing over 25% of the time on the pitch. We’ll expect the sinker to get hit on the ground — but it has, 15.91% of the time, about double the ground ball rate of his other pitches. And we’ll count on him hitting the corners more often than not, but he’s done that too, throwing his fastballs for balls under 40% of the time, about 30% for the changeup, and about 42% on the curve, which is thrown down on purpose. Hellickson is doing all of the things that has made him successful in the past, but he’s getting hit hard, and an increase in ground ball rate (a trait he shares with others on the staff) really shouldn’t be hurting him more, what with Nick Ahmed and Chris Owings playing the middle infield for most of his innings.

Hellickson has continued to hammer away at his favorite part of the zone, down to his arm side both to lefties and righties. His slugging percentage there has not been pretty, though (ESPN Stats & Info). It’s just not working.

Not really sure where the team goes from here with Hellickson, as Jeff and I talked about at length on the last episode of The Pool Shot. Coming into the season, it sure seemed like some of his tried and true techniques were being met with worsening results, especially with respect to the curve and changeup. DRA gives us another indication, however, that continuing to do what he’s been doing is not a promising option.

Tagged with:

One Response to Deserved Run Average and Jeremy Hellickson, As Good As Advertised

  1. […] have had a very good bullpen, especially in contrast to the rotation. And Jeremy Hellickson‘s tricks aren’t working, and Josh Collmenter‘s seem to have a shelf life, and Rubby De La Rosa and his fastballs have […]

Leave a Reply

Your email address will not be published.