Diagnosing Jon Gray

In a fairly surprising turn of events, the Rockies demoted Jon Gray Saturday. Gray has arguably been baseball’s most enigmatic pitcher this year, posting a career-worst 5.77 ERA supported by career-best peripherals — e.g., a 13.4% swinging strike rate (SwStr%) underpinning a 28.9% strikeout rate (K%), and fielding independent metrics of 2.78 xFIP, 3.08 FIP, and 3.15 SIERA. Given our most basic sabermetric understandings of baseball, Gray should be a very good pitcher, even if he pitches half his starts at hitters’ paradise Coors Field.

I have written about how a common-breed Rockies pitcher’s peripherals might be penalized for calling Coors Field home (Gray inspired this bit of research as well). FIP metrics generally underestimate ERA by anywhere from 0.8 to 1.3 runs for home starts (compared to 0.0 to 0.2 runs for road starts), suggesting that Rockies pitchers may underperform (a) their FIPs by 0.35 runs or (b) their SIERAs by 0.65 runs — given error bars, maybe more.

Still, that doesn’t explain why Gray’s ERA is nearly 6 right now. I shed light on the ridiculousness of the move; his strand rate (LOB%) is suppressed and his batting average on balls in play (BABIP) is elevated, even compared to his uniquely bad baselines. I’m not sure there’s much more to it.

Nick Mariano of RotoBaller noted here that Gray’s fastball has been incredibly hittable since his debut and especially this year. Despite my thoughts on the inevitability of regression in Gray’s favor, I wanted to pursue Mariano’s train of thought a little further. Gray’s fastball is bad, but how bad? And why?

Answer: it’s bad. I mean, this probably won’t surprise you, but among 125 pitchers with at least 300 batted balls against their four-seamers from 2015 through 2017, Gray’s fastball ranks 123rd in BABIP allowed (.362) and 106th in line drive rate (LD%) allowed (30.3%). Among 41 fastballs that allowed at least 500 batted balls (Gray’s incurred 505 from 2015-17), Gray’s predictably ranks last in BABIP and 36th in LD%. I’ve been talking about everything pre-2018, but this year has only been worse: 31% line drives (according to Baseball Prospectus’ PITCHf/x leaderboards) and a .406 BABIP allowed (source), respectively ranking 55th and dead last among 73 fastballs thrown 500 times this year. Four-seamers are particularly vulnerable pitches; Gray’s, exceedingly so. It’s hard to believe this is anything less than a trend.

I qualitatively compared the composition of Gray’s fastball to those of Hector Santiago, Ian Kennedy, and Matt Moore — all decidedly unimpressive pitchers who, nonetheless, suppress BABIP against their fastballs. Brooks Baseball has a “grooved pitches” variable; I expected Gray’s to be exorbitantly high relative to his colleagues. Alas, they all float in the 5% to 7% range, showing no significant differences. I looked at pitch location and movement, hoping but failing to eyeball a glaring outlier. There might be some merit to attributing the issue to Gray’s release points, which differ dramatically for his four-seamer compared to his other pitchers, which are clustered at a different release point. But the trend doesn’t hold in 2014-15, and his fastball sucked then, too. I took the quantitative route and ran a few regressions that turned up empty as well.

It’s an unsatisfying conclusion, but maybe it’s just an extremely bad pitch. (There’s a good quote from Ben Lindbergh’s piece for The Ringer this morning about Gray missing middle-middle too often.) You know who else throws an extremely bad primary pitch? Michael Pineda, who has the 2nd-worst BABIP the last four years (min. 400 innings) and whose cutter has incurred a .341 BABIP on more than 900 balls in play. We have all but accepted that Pineda is who he is. We might have to do the same with Gray — that these problems are not simply the product of Coors Field (he has a career 5.02 ERA on the road).

All of these problems begin to compound, too. Although I just wrote off Coors Field as being the problem with his fastball, it is the problem with just about everything else, being an extreme hitters’ park that, contrary to popular belief, plays up BABIP way more than it does home runs. Higher BABIP leads to more baserunners leads to lower strand rates (r² = 0.28) leads to more runs allowed. The fielding independent metrics, founded upon regression analysis, look at key variables (strikeouts, walks, ground balls, etc.) and inherently assume all else (namely, BABIP and LOB%) is held constant.

In other words, FIP and xFIP and SIERA evaluate a specific set of pitcher outcomes and assume everything else is league-average, for better or, in Gray’s case, for worse. Since 2000, Rockies pitchers have underperformed FIP by 0.32 runs, xFIP by 0.53 runs, and SIERA by 0.54 runs (very slightly more conservative than the numbers from my ERA-minus-SIERA post linked at the beginning). Those deficits encompass more than 26,000 innings thrown both home and away. (I know you’re expecting splits: 5.08 ERA versus 4.23 xFIP at home, 4.50 ERA versus 4.41 xFIP away.) We can’t reasonably expect Gray to come within half a run of his SIERA in the long run as a true-talent league-average BABIP pitcher, which he is not. It appears he’s worse than that. We should adjust our expectations accordingly. Gray’s career ERA-minus-SIERA differential stands at 1.10 runs. I think it’s a bit exaggerated, but it may not improve by a substantial margin. This says nothing of his potential issues while pitching with runners on base, as Lindbergh discusses. However, I’m not completely sold on this being a legitimate issue; it’s supposedly the same problem that affected Robbie Ray in 2016 with his historically high BABIP allowed, only to turn around and have a stellar 2017 campaign devoid of such problems.

This brings us full circle: Gray can be both (1) not as good as we hope and also (2) “due” for regression. Those tenets are not mutually exclusive. Gray’s 2.62 ERA-minus-SIERA screams that he’s much better than his 5.77 ERA suggests, irrespective of Coors Field or his fastball. He will come back from AAA, cash in on some good luck on his BABIP and LOB%, and have a dominant second half, and we will attribute the success to his demotion. Fact of the matter is pitchers historically have rarely, bordering on never, sustained a .386 BABIP and a 63.1% strand rates for a full season.

Granted, the folks reading this post are more likely Gray believers than dissenters. I don’t need to convince you he’s better than this. But I do need to convince you Gray’s 2.78 xFIP betrays you. Gray’s absolutely a regression candidate, but he’s still not the staff ace the Rockies have so desperately needed for so long, nor is he the fantasy No. 2 you see under the surface. Given his current peripherals, he’s more likely a true-talent 4.00 ERA guy; given his career marks, maybe closer to 4.30 ERA. And if you believe in the issues with men on base: even worse.