After starting his career with a 2.88 ERA in 491.0 innings pitched from 2013 to 2015, Sonny Gray proceeded to post a 4.59 ERA over his next 409.2 innings, spanning from 2016 to 2018. His career was looking like it was starting to trend toward mediocrity, but then Gray came back revitalized in 2019 with a 2.87 ERA and solid peripherals to boot. He posted a career-best 19.3 K-BB%, and thus his lowest FIP ever in a full season. Unfortunately, I think that it is, in part, a mirage.
Some of the regression red flags are more obvious. I doubt there are any full believers in Gray’s 2.87 ERA as opposed to his 3.42 FIP or 3.97 SIERA, and his HOTEL spells trouble, too. His 0.87 HR/9 is fine, but HR/9 is noisy, and his .255 BABIP is 25 points down from his career .280, and his 79.7 LOB% is a career-high. Of pitchers with 200 or more batted balls, Gray’s -.030 batting average minus expected batting average ranked in the 82nd percentile, meaning he was more fortunate than 82% of pitchers. His wOBA-xwOBA of -.052 ranked in the 83rd percentile, too, so Gray has either had good fortune due to his defense, park, or luck (or all three!).
FanGraphs’ Alex Chamberlain has done some really interesting work looking at xwOBAcon by pitch type. Essentially, what Chamberlain is arguing is that we should be looking at pitchers by their pitches individually since looking at overall xwOBA and xwOBAcon can be deceiving. In looking at pitch-type xwOBAcon, we are able to tease out specific pitches that have overperformed (or underperformed) erroneously. He uses Zack Wheeler as an example in his article, but what we are essentially looking for are pitches on the extremes—with an xwOBAcon inflated or deflated significantly—and mentally regressing their performance from there.
Surely there are pitchers who naturally will hold high xwOBAcon numbers—think Chris Archer or Nick Pivetta—and there are also pitchers with low xwOBAcon pitches (e.g., Dallas Keuchel or Kyle Hendricks). The rationale is that limiting hard contact sustainably is incredibly difficult, even though certain pitch properties influence exit velocity and launch angle (which is somewhat intuitive). Because it is easier to estimate strikeouts and walks year to year, it makes sense that these should hold more weight than mere statistical noise (as often happens with xwOBAcon on an individual pitch basis). Chamberlain explains it a lot better than I do, but essentially what we’re going to look for is a disparity in Gray’s 2019 numbers from both league averages as well as his own career averages.
While Gray has historically been a player to limit hard contact—and thus keep his relatively xwOBAcon low—his numbers are misleading. Statcast knows this, to an extent. Given its inputs—strikeouts, walks, and batted balls—it knows that Gray is not the .265 wOBA pitcher that he was in 2019. Let’s compare his wOBA on contact and xwOBA on contact (xwOBAcon), by pitch type:
|wOBA (contact)||xwOBA (contact)|
For the most part, everything is in order, but Gray was quite fortunate on batted balls with both his curveball and slider—this is the bulk of the discrepancy in his wOBA and xwOBA. However, there is more cause for concern. Let’s compare his xwOBAcon to both his career averages, and league averages.
First, let’s compare Gray to the league in 2019:
|Gray (xwOBAcon)||League (xwOBAcon)|
In the first table, it showed that, given the batted-ball quality, Gray’s wOBA was significantly better than it should have been on his curveball and slider. Here, we looked specifically at xwOBAcon and we see that both his slider and curveball xwOBAcon are significantly better than the league average. While highly unlikely (due to the extent that they were better than the league averages), there is a possibility that this is sustainable.
Gray’s 2015-2018 (Statcast era) to his 2019:
|xwOBAcon (2019)||xwOBAcon (2015-18)|
Granted, Gray’s xwOBAcon was elevated on both of his fastballs in 2019, so perhaps he’s underperforming on these particular pitches, but given that they’re similar to league average for both respective pitches, I’m inclined to believe that they’re right around where they should be at. Again, the focal point here is Gray’s breaking pitches, especially because this is the first time in his career—and it’s not even close—that Gray threw more breaking pitches than fastballs.
What’s interesting is that, sure, he changed the shape of his slider, and it seems like he’s sequencing his pitches better, but he isn’t getting more whiffs on his breaking pitches. Oddly enough, the pitch that took a big step forward in 2019 for him as a swing-and-miss pitch was his sinker, which is seeing its demise across the league. This is likely a direct result of his slider becoming a more side-to-side pitch, paired with his increase in its usage. Sinkers and sliders naturally play well off one another because they have the ability to tunnel together and then “repel” off one another. His sinker had a career-low in zone-swing percentage, which tells me that it is more deceptive than it has been in the past due to its pairing with his reshaped slider.
Given that his four-seam fastball is a mediocre pitch, there’s a chance to he can scrap it altogether, but what I think is more important is that he uses his sinker in favor of it. It gets more whiffs and significantly more ground balls, and has historically been a superior pitch to his four-seam fastball, in terms of wOBA, xwOBA, and ground-ball percentage, so perhaps he should lean on it more. In any case, even if he may have a somewhat improved fastball and a better pitch mix, there don’t appear to be any dramatic improvements.
Perhaps, then, this will help explain this:
— Alex Chamberlain (@DolphHauldhagen) November 25, 2019
It’s not out yet, but Chamberlain’s Tableau (which sounds like a thought experiment but is not) is set to feature even more interesting things. Although it’s highlighted, I’m not looking to single out Brandon Woodruff. If you look near the bottom of the list, Gray makes an appearance as one of the biggest overperformers in strikeout percentage minus expected strikeout percentage. His 28.9 K% should, according to this table, be something more like 26.6%. Steamer currently projects about a 25.7 K% in 2020, so perhaps ~26% is what we should expect going forward. If we take his xK%-xBB% and compare it to other starters’ K-BB%, Gray looks more like Tanner Roark or Miles Mikolas than Trevor Bauer. Overall, the Tableau has him at more of a true .299 xwOBA, and that’s not accounting for impending xwOBAcon regression (which is likely but not definite).
Sonny Gray has made good changes, but at the end of the day, his most frequently used pitch is still a bad fastball. He’s made some necessary changes, but the next step might be to further alter his pitch mix. He should continue to lean on his breaking pitches—which I think are legitimately improved—but he would be best suited continuing to throw his slider a lot, and perhaps the biggest change he can make is to ditch his four-seam fastball for his sinker. His fastball has elite spin, but given that it has little arm-side movement (and he doesn’t throw it at the top of the zone), it’s not going to be a good pitch for him. In any case, his slider and curveball are nearly certain to show some regression next year, which means that if no changes are made, Gray is much more likely to look average or above-average than elite.
(Photo by Brian Rothmuller/Icon Sportswire)