Drew Hutchison has a real problem from the stretch

Coming into this season, Drew Hutchison faced some high expectations following a solid 2014 in his first season coming back from Tommy John surgery. A combination of very good peripherals, the excellent slider he flashed late last season, and the fact that he was not yet 25 led many, myself included, to think he was a potential breakout candidate, potentially as good as a #2 starter (top 50 or so in MLB). Just over halfway through the 2015 season, it's safe to say that hasn't come to fruition, with his command and sharpness of stuff having been inconsistent and Hutchison sporting an unseemly 5.33 ERA despite a very solid 3.65 FIP.

And therein lies the rub with Hutchison, 61 starts and 341 innings into his career, the equivalent of almost two seasons. What one thinks of Hutchison really depends on what one looks at. Based on his peripherals, he's been quite good, with a 3.90 career FIP (99 FIP- or almost perfectly league average), resulting in 4.4 fWAR that puts him as a mid-rotation starter. However, his ERA is 85 points higher at 4.75 (119 ERA-), with only 2.2 RA-9 WAR and 1.3 bWAR. So purely measured by results at preventing runs he's merely been a backend starter and this year right around replacement level.

With the Blue Jays in dire need of quality starting pitching, the big question is: will the real Drew Hutchison please stand up?

As a general rule of thumb, when ERA and FIP diverge in the way they have for Hutchison, FIP generally is a better predictor of the future than ERA (see here and here for more detail). 341 innings, though not a small sample, is not a big enough sample that weird randon variation can't make a pitcher look a lot better or worse than he actually. As an example, for the first 400 innings of his MLB career from 2010-12, Jeremy Hellickson posted an excellent 3.06 ERA despite mediocre peripherals and a 4.46 FIP. That was accomplished on the back of .244 BABIP, the lowest in the MLB and basically impossible to replicate, at least to that degree.

Since then, Hellickson hasn't had nearly the same results, posting a 5.01 ERA and .311 BABIP despite actually improving his underlying peripherals. He's had some health issues and will likely be better than he has been the last couple years, but his talent was never that of a 3.00 ERA pitcher despite putting up those results for 400 innings. Likewise, there's plenty of reason to expect Hutchison's future results, including the second half of 2015, to look a more like the average to slightly above pitcher his underlying numbers suggest he is, which frankly is all the Jays need.

Nonetheless, 341 innings is not an insignificant sample either, and the reality is some pitchers do have persistent deviations between peripherals and results that do reflect real strengths or weaknesses. A good Bayesian is always updating expectations from real data, so it's worth looking deeper to see if there's a real issue. As it turns out, there is a real reason to be concerned.

The first place to look, as with Hellickson, as at BABIP. At .358, Hutchison has the highest BABIP of any qualified starter in 2015. However, in previous years it's been right in line with league average in the low-.290s. His line drives are way up, and that's consistent with spotty command resulting in batters squaring him up, but his hard contact are in line and his soft contact numbers are actually up. It's not like he's getting crushed, and given the context, there's little reason to thing it's more than a blip. That's not to say he hasn't earned the higher BABIP - to some extent he has - but there should be a very strong dose of regression.

The second strong possibility is splits between when the bases are empty and runners on, because some pitchers are significantly worse out of the stretch. And here we see the crux of his struggles. Consider his career splits for various statistics:

With runners on, Hutchison strikes out 6% fewer batters, walks almost 3% more batters, and is significantly more home run prone despite giving up a similar batted ball profile. With the bases empty, he's a borderline ace and with runners on he's a borderline scrub. Granted this is not unique to Hutchison, as most pitchers aren't as good from the stretch and league wide there is a significant split. But Hutchison goes well beyond that. Below, I've shown him relative to the 2014 MLB average for SP in K%, BB%, and HR/9 for the two splits:

Hutchison goes from striking out 16% more batters with none on than other MLB starters to below average with runners on. His walks go from better than average to worse too. Most significantly, instead of a small HR problem due to his fly ball profile, he has a big one with runners on (which is the last ting you want). MLB pitchers in generally actually allow about 10% fewer HR with runners on.

Is this possibly just random variation? Almost certainly not for the strikeouts and walks at least, as they are more stable stats that are very reliable when talking about samples over 500 PA, which is case for Hutchison in both splits. Moreover, on a year to year basis the splits have been there, it's not just one bad year skewing otherwise decent splits which might be grounds to think it's more likely to be random. It's been an ongoing issue.

That doesn't answer the more important question of why this is the case. Less velo? Less break? Worse control or command? To look at a more granular level, I downloaded all this PITCHf/x data from Baseball-Savant. Unfortunately, the raw output doesn't indicate which pitches happen with bases empty and men on. So had to use an algorithm to approximately sort things out. It's not perfect, and I omitted a little bit of the data, but I'm confident it's well over 95% correctly sorted.

On the very positive side, the raw stuff is not worse out of the stretch, as his average velo is the same for all pitches: 92.0 MPH with the bases empty v.. 91.9 MPH with men on for his fastball, 84.5 vs. 84.3 on his slider, 85.2 v. 85.2 on his change-up. The same is true for movement in all direction, break, and spin. The only exception is on his slider he gets more spin in the stretch (715 RPM vs. 656); I don't know if that's material, but it certainly wouldn't come close to explains things in any event.

Instead, it appears the problem mostly lies in control. With the bases empty, Hutchison is roughly league average in terms of his ball/called strike ratio at 36%/18%, and consequently how often he's ahead or behind. He's also roughly league average in terms of how often he gets into dominant and very dominant hitter and pitcher counts.

However, with runners on his control numbers are markedly worse at 37%/14% balls/called strike overall. He maintains his swinging strike rate, so it means more fouls and balls in play. On the first pitch, it's even more stark, at 43%/26% compared to 39%/33% otherwise. So he's behind more often, hitters are in better counts where they can selectively sit on pitches, and he's gets fewer dominant pitcher counts where he can put batters away.

There's also the possibility it's a command problem as in some sense command/control are flip sides of the same coin, but I'm less inclined to think so. Poor command usually leads to contact problems, but the data are at best amibguous. His batted ball profile and BABIP are similar, with the big difference being HR/FB. That is a notoriously noisy stat, and so could reflect random variation that will regress. His hard/medium/soft FanGraphs data is basically the same, though I'm loath to read too much into that since it's so new.

But Hutchison definitely appears to have a significant control problem out of the stretch, and it's hurting his results. I don't if there's something mechanically or in his approach, that's not something in which I have any expertise. but it's not addressed than he can be expected to underperform his stellar peripherals, if not to the same extent as in the first half. Right now though, it's holding him back from realizing his potential as a mid-rotation starter or perhaps better.