The Most Predictable Man in Baseball by Ben Clemens June 14, 2019 The Tampa Bay Rays are having a tremendous year so far, better than anyone could have expected. They’re a half game out of first in the perennially difficult AL East, and that might be underselling how good they’ve been this year — their BaseRuns record is the best in baseball. How have they done it? Their pitching staff has been the best in baseball by a huge margin, posting a 3.02 ERA and a 3.34 FIP, both of which are miles better than second place. The hitting has been good, but pitching has the Rays playing like a championship contender. That pitching staff has been a many-headed monster this year, and Yonny Chirinos has been a key part of it. He’s bounced back and forth between starting and following an opener (headlining?) over 75 innings of work, compiling a 2.88 ERA and 4.05 FIP in his second major league season. He was above average last year as well — a matching 3.51 ERA and FIP over nearly 90 innings. He sports a 21.5% strikeout rate and a sterling 4.9% walk rate. In short, Chirinos looks like a mid-rotation major league starter for the foreseeable future. What’s truly amazing about him, however, is that he’s doing that while being the most predictable pitcher in all of baseball. If you’re behind in the count against Yonny Chirinos, it’s going to be a long day for you. His splitter, which he only learned in 2017, is lights-out. It’s been the third-most-valuable splitter in baseball this year, behind relievers Hector Neris and Kirby Yates. It generates truly video game numbers: a 45% whiff rate, 2.5 ground balls for every fly ball, and a .155 wOBA on plate appearances that end with a splitter. When Chirinos has the advantage, he’s not shy about going to the split: he throws it 43% of the time, more than twice as often as his overall rate of splitters. No, if you want to beat Chirinos, you need to avoid the splitter. If you end up in a two-strike count, you’ll probably wave at air before heading back to the bench. Get ahead in the count, however, and things change. Chirinos has an effective fastball, a 94-mph sinker with huge horizontal break that runs in on the hands of righties. Still, it’s a fastball, not a world-destroying offspeed pitch. There’s no question which offering you’d rather face. Why do I keep referencing the count when it comes to which version of Chirinos a batter will face? Well, consider this: when he’s behind in the count, Chirinos throws 84% fastballs. He’s not sinker-only, but he’s close to it — he’s thrown only two four-seam fastballs while down in the count all year. It’s not quite accurate to say that Chirinos is a one-pitch pitcher when behind in the count, but it’s close enough as a first approximation. That 84% fastball rate when behind in the count doesn’t make Chirinos unique. Sean Doolittle has thrown only fastballs when behind in the count. Zack Britton is close behind — he’s throwing 99% fastballs to get back into at-bats. Among pitchers who have thrown 100 pitches while behind in the count, Chirinos has the 21st-highest fastball rate. The top two places on the list aren’t all that illustrative. Maybe Doolittle throws 100% fastballs when he’s behind, but he throws 92% fastballs overall. No matter the count, the game is the same: he’s throwing a fastball, and the hitter needs to find a way to deal with that fastball. Chirinos, on the other hand, isn’t a fastball-only pitcher, or even someone who throws a ton of fastballs. His 57.1% fastball rate is barely in the top third of all qualified starters this year. Not only does he throw more than 20% splitters, but he has a cromulent slider too. He’s a true three-pitch pitcher — at least, in his overall stats. To get a sense of how bifurcated Chirinos’s pitch selection is, think of it this way: When he’s behind in the count, Chirinos throws the 11th-most fastballs in baseball. When he’s ahead in the count, he throws the 25th-fewest (out of 282 pitchers). No pitcher in baseball has a larger gap between fastball rates when ahead and behind. I already mentioned that Chirinos throws 84.3% fastballs when he’s behind. When he’s ahead, that number dips to 33.7%. That 50.6% gap is the most in baseball by a lot. To borrow an old Jeff Sullivan descriptor, the gap between Chirinos and second-place Seranthony Dominguez is the same as the gap between second and 10th place. Here’s that list: Behind? Throw a Fastball Player FB% Increase When Behind Yonny Chirinos 50.6% Seranthony Dominguez 43.7% Tyler Glasnow 42.2% Ryne Stanek 41.3% Victor Alcantara 40.2% Brad Hand 39.3% Spencer Turnbull 38.9% German Marquez 38.6% Trevor Bauer 36.7% Brandon Kintzler 36.4% Now, pitch mix isn’t as easy as just saying you should be unpredictable and throw every pitch in equal proportion in any count. The rule of thumb that you should throw fastballs when behind and secondary pitches when ahead is logical: balls cost you more when you’re already down in the count. What the batter expects matters too — fastballs work a lot better when batters aren’t expecting them. There’s no way around the fact that no one rule of thumb works for every pitcher. Hitter expectations, command, and type of secondary pitches all affect the optimal mix, and there’s no guarantee it’s stationary. Still, 50% sounds like a lot to increase your fastball rate by. When I looked into Trevor Bauer’s fastballing ways earlier this year, I came up with a method of modeling that I’m pretty happy with for looking at whether pitch selection in a given count makes sense. Looking directly at results runs into sample size issues — a single extra-base hit against a Chirinos off-speed pitch would have an outsize effect given their rarity in some counts. Instead, the idea is to look at some basic outcomes of a pitch — called strikes, balls, swinging strikes, foul balls, and balls in play. There’s some argument that fouls aren’t stable enough to include in analysis like this, and I’m honestly not sure which side of that argument I come down on: foul ball rates are certainly volatile, but there’s probably some skill there. In the following analysis, I’m treating foul balls as in the pitcher’s control, but if you want to regress those to league average by pitch type, I could totally see the logic in that as well. Swinging strikes, called strikes, and balls are all pretty straightforward — pitchers have a ton of agency in all three. The last thing we need to work out in terms of what value Chirinos is getting from throwing so many fastballs in fastball counts is the value of batted balls. When doing this analysis with more established pitchers, I like to separate out in-zone contact and out-of-zone contact. After all, making contact with a wipeout slider just before it hits the dirt is a lot different than hitting a slider that catches the middle of the strike zone. In Chirinos’ case, however, there simply isn’t enough data to split things out that way. Opponents have only put 27 out-of-zone splitters into play in his career, for example. The sample size is just too small. I considered a few options (normalize Chirinos’ in-zone results by some constant I figured out from all splitters, try to build a model that uses a few factors about pitchers to predict out-of-zone rates, plus some not-very-well-thought-out other attempts), but I settled on a more straightforward approach: ignoring in-zone versus out-of-zone and just taking the xwOBA on contact of every ball in play Chirinos has allowed on each pitch. With more data, I might try to split things out more (such as looking at fastball contact in specific counts or situations), but the data is a bit too sparse to do that until he pitches more. Is it perfect? Definitely not. Does it get the job done? I think so. On to the results! Chirinos throws an absolute ton of strikes when he’s down in the count. His fastball is in the zone 67% of the time — he’s simply not willing to give up a walk. Batters are no dopes, though — they swing at 75% of those in-zone fastballs. Put it together, and sprinkle in a Markov chain-based accounting of the value of strikes, balls, and fouls, and here’s what it looks like: Change in wOBA, Fastball, Down in Count Result Change(wOBA) Frequency Ball 0.136 24.4% Strike -0.099 24.9% Foul -0.064 22.2% In Play 0.053 28.5% Total 0.010 100.0% Not great, Yonny. Throwing a fastball leads to better outcomes for batters by 10 points of wOBA. I can’t stress enough that this is only a toy, and that there are huge margins of error on any of my calculations. Still, it seems like the complete lack of whiffs on fastballs isn’t helping. Let’s look at the same data for the splitter: Change in wOBA, Splitter, Down in Count Result Change(wOBA) Frequency Ball 0.136 27.3% Strike -0.099 45.5% Foul -0.064 4.5% In Play -0.004 22.7% Total -0.012 100.0% Now if the fastball sample size was small, this is microscopic. I don’t have a ton of faith in basically any number here. Still, it seems likely that when Chirinos throws a splitter, he’s getting value. Batters take a ton of strikes (they barely swing at half of his in-zone splitters, and he’s been good at spotting it in the zone when he’s down in the count) and whiff a lot when they do swing. Essentially, they’re underprepared for the pitch. Last but not least, let’s look at sliders. That’s another way Chirinos could go if he wanted to throw fewer fastballs in hitters’ counts. If his splitter is a glimpse into how much better Chirinos could be, the slider is a peek at the other side. He hasn’t been as good, thus far in his career, at locating the slider in the zone, and that greatly saps its value. Change in wOBA, Slider, Down in Count Result Change(wOBA) Frequency Ball 0.136 34.4% Strike -0.099 26.7% Foul -0.064 16.7% In Play 0.008 22.2% Total 0.012 100.0% So what do these results tell us? Basically, it’s hard to say! Chirinos is doing himself no favors by throwing so many fastballs. With very few strikes and foul balls, he gets few strikeouts, ending up with a ton of balls in play. It shows in his stats thus far — splits are always a little dicey from a sample size perspective, but Chirinos has struck out only 12.5% of batters after a 1-0 count for his career, as compared to 18% for the league as a whole. For completeness’ sake, here are the pitch values when Chirinos is ahead in the count. If there’s anything interesting here, it’s that one piece of baseball wisdom absolutely checks out: if you, the pitcher, are ahead in the count, you’d prefer to throw a ball rather than have your opponent put the ball in play. Change in wOBA, Fastball, Ahead in Count Result Change(wOBA) Frequency Ball 0.034 40.1% Strike -0.141 18.7% Foul -0.033 20.1% In Play 0.187 21.1% Total 0.020 100.0% Change in wOBA, Splitter, Ahead in Count Result Change(wOBA) Frequency Ball 0.034 42.0% Strike -0.141 26.8% Foul -0.033 14.8% In Play 0.130 16.3% Total -0.007 100.0% Change in wOBA, Slider, Ahead in Count Result Change(wOBA) Frequency Ball 0.034 31.9% Strike -0.141 25.8% Foul -0.033 20.3% In Play 0.142 22.0% Total -0.001 100.0% Do I think this analysis is groundbreaking and that I’ve solved something previously unknown about Yonny Chirinos’s game? Honestly, no. The Rays are really good at analyzing baseball. They’ve no doubt thought about this, and they’re better at massaging the numbers to get useful data. I think it’s notable though. Yonny Chirinos throws way more fastballs when he’s behind, and maybe he throws a few too many. When he’s ahead, however, he throws an entire mountain of splitters, and he should maybe throw even more. Not enough analysis has been done about how you should change your pitching style based on count. It’s an interesting baseball problem, an interesting game theory problem, and really just a tremendous mind game. The case of Yonny Chirinos provides an interesting window into what might make sense. Throwing more fastballs when you need strikes isn’t necessarily bad, and it isn’t necessarily good. It’s contextual, to the point where we can look at an extreme outlier and not really know whether being so extreme is good or bad. The next time I watch Chirinos pitch, I’m going to be watching every 1-0 pitch intently, waiting to see yet another sinker and yet another batter expecting a sinker. Even better, maybe I’ll see a splitter! Rarity is beautiful, after all.