Clayton Kershaw, Missing the Zone
Here, take a look at Clayton Kershaw throwing a 3-0 pitch in 2018:
Pretty straightforward, huh? Okay, now take a look at the same situation in 2019:
Can you spot the difference between the two? I’ll save you some time — I can’t either.
Not very interesting, right? Well, that’s not really what we’re here for. You see, Kershaw’s behavior when he throws a strike on 3-0 doesn’t look very different. In 2019, however, his overall 3-0 strike-throwing changed greatly. Was it real? Was it a fluke? Does it tell us something deeper about Kershaw? Let’s investigate.
Take a look at Kershaw’s 3-0 zone rate in each year of his career, as well as the league average for that year:
| Year | Kershaw | League |
|---|---|---|
| 2008 | 46.2% | 53.1% |
| 2009 | 55.6% | 54.4% |
| 2010 | 50.0% | 54.1% |
| 2011 | 65.8% | 54.0% |
| 2012 | 66.7% | 54.9% |
| 2013 | 70.8% | 55.2% |
| 2014 | 73.7% | 55.1% |
| 2015 | 63.6% | 56.1% |
| 2016 | 83.3% | 55.7% |
| 2017 | 83.3% | 62.4% |
| 2018 | 71.4% | 61.3% |
| 2019 | 33.3% | 61.0% |
Is this a small sample? Absolutely! Kershaw faced only 18 3-0 counts last year. But while the league has only gotten more aggressive when it comes to flooding the zone on 3-0, his rate has gone directly the opposite way.
Is this a planned change? Is he missing the zone by inches? Heck, is the zone miscalibrated on Baseball Savant? Maybe the “out of zone” pitches are still being called strikes — it’s not like that never happens. The only problem with that theory is that it’s not true. Kershaw missed the strike zone 12 times on 3-0 last year. Ten resulted in a take and a walk; on another two, the batter swung and fouled the pitch off.
Was he getting squeezed by the umps? Not as far as I can tell. Here’s the closest call of those 10 walks:
That’s a ball. Nothing weird is happening with the classification; Kershaw is just missing more. Remember, as well, that the rest of the league gets classified with the same strike zone Kershaw does, and the overall zone rate is only headed higher. He’s simply lost the zone.
Now that you know this weird factoid, it’s time for the logical next step — using this small detail to make a great logical leap and reach a sweeping conclusion about Kershaw. There’s just one problem. I have absolutely no idea what that conclusion should be.
Consider: most pitchers who miss the strike zone frequently on 3-0 counts do it because they’re wild. Here’s the list of the 11 pitchers with the lowest 3-0 zone rate over at least 10 pitches in 2019, as well as their overall walk rates:
| Player | 3-0 Zone Rate | Walk Rate |
|---|---|---|
| Wily Peralta | 18.2% | 10.8% |
| Luke Jackson | 20.0% | 8.3% |
| Clay Holmes | 25.0% | 15.0% |
| Andrew Miller | 27.3% | 11.4% |
| Tyler Kinley | 30.0% | 16.3% |
| Justin Wilson | 30.0% | 11.5% |
| Sean Reid-Foley | 30.0% | 14.0% |
| Héctor Neris | 30.8% | 8.7% |
| Scott Barlow | 30.8% | 11.9% |
| Kyle Crick | 30.8% | 15.5% |
| Clayton Kershaw | 33.3% | 5.8% |
No one else on the list has a walk rate below 8%. Kershaw remains one of the game’s premier control pitchers; his 5.8% walk rate is his worst mark since 2012 and still in the top 20% of starters. No, this isn’t a case of Kershaw spiraling out of control, turning into Mitch Williams on the mound. It’s merely a specific place where he suddenly struggles to throw strikes.
When you’re dealing with an 18-pitch sample, it’s good practice to be skeptical. Huge samples can blot out contextual differences; over a full season, the quality of opposing batters tends to be relatively close to league average, for example, but in a one-PA sample, sometimes you’re facing Mike Trout. Was Kershaw simply responding to specific game situations?
Not really. Of the 12 out-of-zone pitches he threw on 3-0, nine were in situations where a walk was particularly valuable for the batter relative to a hit — with the bases empty and less than two outs, in a game where the score was so lopsided baserunners were of utmost importance, or to put a runner on third base with only one out — similar to the proportion of these situations he faced in previous years. And while the hitters were slightly above average when it came to production on contact, it wasn’t by a significant amount; they produced roughly like Avisaíl García (who was one of the 3-0 walkers) when putting the ball in play in 2019, hardly a fearsome comparison.
It’s not a pitch selection issue. As Travis Sawchik and Jeff Zimmerman have previously noted, Kershaw is a predictable type; fastballs when he’s behind, breaking balls when he’s ahead, and little deviation from the “naive” pitch option. In fact, all 18 of the 3-0 pitches in our sample were fastballs. Kershaw isn’t branching out into new experiences and seeing the costs of throwing secondaries when behind; he’s simply missing with his fastball.
Do 3-0 walks even mean anything? After a 3-0 count in 2019, batters produced a gaudy .314/.725/.555 slash line last year, good for a .544 wOBA. Kershaw’s 19 batters (one of his 3-0 plate appearances appears not to have been tracked by Savant) produced a .167/.737/.167 line, good for a .518 wOBA. The sample is far too small for this to mean anything, but despite his garish zone rate, 3-0 counts didn’t hurt him in 2019.
Why is that? Shouldn’t the free walks hurt so much that they overwhelm everything else? In the long run, they probably will. His 3-0 stats aren’t sustainable. A .250 BABIP, a zero ISO; those aren’t things that any pitcher can credibly create. The league average after 3-0 counts last year was a .340 BABIP and a .241 ISO — hitters do more damage on contact when they’re ahead in the count, not less. Even Kershaw, for many years the best pitcher in baseball, has allowed a .330 BABIP after 3-0 counts, though with an impressively low .093 ISO, for his career as a whole. That sample stretches back to 2008, so it’s not directly comparable with the 2019 league-wide metrics, but the point remains: Kershaw needed some luck to achieve average results after 3-0 counts last year.
Did we actually learn anything from this? Most likely not. But the odds of this being truly random are actually quite low. If we assume that Kershaw’s “true” 3-0 zone rate is his career rate before 2019, 61.4%, the odds of throwing six or fewer strikes in 18 chances are a mere 1.5%. If we assume that his true talent rate is instead the 70.4% rate he showed from 2012 to 2018, the odds are an even slimmer 1.2%. Even if his true talent rate were 50%, he’d hit 33.3% or lower only 12% of the time. A rate this low is probably not purely the result of chance.
There’s something interesting there, hidden inside this nonsense stat about literally only 18 pitches in Clayton Kershaw’s extensive career. When it comes to baseball statistics, our brains have been conditioned to ignore small samples. It took many of us years to get there; those darn batter-vs.-pitcher matchup statistics baseball broadcasts showed us as children take time to forget. And now that we’ve leveled up our understanding of statistics, we dismiss any small sample out of hand. Hah! 18! Come back to me when you have 1,000!
Not all small samples are meaningless, however. Something as simple as zone rate should stabilize reasonably quickly. If a pitcher had a stationary true talent zone rate, his observed zone rate would converge on that number with great haste. Think of it in this naive way; take the 50% zone rate from before and make it a coin. Flip the coin 100 times. How often do you achieve 40 or fewer heads? Only 2.8% of the time. Heck, over 50 flips, you see 20 or fewer heads only 10% of the time. In truly random processes with a stable success rate, outliers go away in no time at all.
Does this mean we should be able to predict pitcher performance? Nope! What it means is that pitcher true talent is constantly changing, likely unpredictably. One day, a pitcher might not be able to find the zone. The next, he might be able to paint the corner with his eyes closed.
That might sound similar to a random process with a static true-talent mean, and for our modeling purposes, it might as well be. But it’s different. In baseball, Kershaw flips mostly tails sometimes, and there’s nothing weird about that. Maybe he’s choosing to flip tails. Maybe heads got the yips. Maybe his coin is sometimes 30% tails and sometimes 70% tails. Anything is possible.
What can we do with this knowledge? I’m not sure yet. Possibly nothing. Probably nothing. It’s tremendously unlikely, to the point of absurdity, that Kershaw had the same strike-throwing talent on 3-0 counts in 2019 that he did in his career until then. But without further research, we can’t say much, if anything, about what that means for 2020. For that, we’d need more data, a better study; do pitchers whose observed zone rates plummet tend to recover? Do they stay low?
Who knows? I certainly don’t! That’s a study for another day. In the meantime, let me leave you with Clayton Kershaw missing the zone badly on a 3-0 pitch to Ketel Marte. Sometimes, balls are just happy accidents:
Ben is a writer at FanGraphs. He can be found on Bluesky @benclemens.
Kershaw’s idea is to not throw any meatballs, even on 3-0. So, if you walk, you walk. He’ll get the next guy.
If so, it’s a new tactic he adopted in 2019. The very first gif is as much of a meatball as any pitch from Kersh can possibly ever be.
I can kind of squint my eyes and see some logic in this… (1) If the outcome is most likely going to be a walk anyway, he kind of just gets it over with. (2) If a batter is swinging regardless of the pitch, his goal is to not give them something that could potentially be hit well, regardless if it ends in a walk. I think the zero ISO is the whole point, not a fluke result.
I understand that hitters have been more aggressive in 3-0 counts over at least the last 2 years, which happens to coincide with Kershaw’s fastball gets more hittable. It’s not outlandish to presume that he might be nibbling a little more than in the past.
I’m a little puzzled at why walks would be considered more valuable relative to a hit in a situation with a lopsided game, could you please expand on that? And what parameter is “so lopsided” exactly? If I was pitching up in a lopsided game I’d probably care less about walking a guy
Yeah sure thing. When I say relative, I’m talking about relative to average; so on average, a single is better than a walk, and extra base hits are a lot better than a walk. But a single is worth the same as a walk with the bases empty, and a double isn’t THAT much better than a walk sometimes. For example, a double is WAY better than a walk with a man on second and two out (it’s roughly a run better) but barely better when leading off an inning. This means that there are situations where if your tradeoff is more walks in exchange for lower likelihood of a hit, you want to make that trade when walks are at their least valuable to the offense, relative to hits — with runners aboard (because hits advance runners, while walks don’t), or with two outs.
In the situation where it’s a lopsided game, the math is basiclally this: the way you lose is with a lot of baserunners. If you’re up by six runs with nine outs left to get, you don’t need many outs per run allowed — avoid crooked numbers, and you win. So free baserunners, with no chance for your fielders to make an out, are the enemy. On the other hand, if you’re up by only a run or two, now extra base hits start being scary. The most *likely* outcome of a given walk is that the runner doesn’t score, even if they increase the likelihood of a lot of runs scoring. So maybe if it’s close, you shouldn’t give the guy something to hit and hope he gets himself out. Generally speaking, the wisdom that you never want to walk a guy with a big lead is just right. That’s an analyst classic that really does stand up to scrutiny.
Ben, I always struggled with this logic as well. So, what you’re saying is, if I hit a single, there’s a chance I could get thrown out rounding the base too far, or that a another baserunner could get thrown out trying to advance. While, with a walk, there is no chance for someone to get thrown out. Is that right?
Something else I thought about was that typically a pitcher throws more pitches when he walks someone than when he gives up a hit. So, giving up walks has a better chance of fatiguing or flustering the pitcher. Does this thought have any merit?
Thanks
As to the first point, not really. That’s true on the margin, since if you’re thrown out going to second you get credit for a single, but that’s mostly not what I mean. If you look at the linear values in wOBA, a single is worth significantly more than a walk; a walk is .690 and a single is .870 in 2019, for example. That’s because there are some situations where a single helps a lot more than a walk, and wOBA’s linear weights are averaged across all base/out states based on the frequency of each state.
I’m just saying that there are situations where a walk and a single are basically the same. But a batted ball that turns into a single sometimes turns into an out, while a walk is automatic. So in situations where a walk is close in value to a single, you shouldn’t walk people. But with runners on second and third, they aren’t close in value, particularly with two outs. wRC+ and wOBA average across all contexts, which is why we call them context-neutral. But in a given PA, the context is never neutral — it’s whatever the current game state is. That’s basically what I’m saying. Not sure if that clears it up, it’s a weird concept to write in words, I’m having trouble getting out what’s in my head, if you know what I mean.
As to the second point, yeah, that’s interesting and probably worth studying. You’d have to do some controlling — worse pitchers walk more batters, and are also, you know, worse — but it sounds worth looking into.
If you’re allowing for the possibility of a single and a TOOTBLAN, you also have to account for a single and an error. My guess is it mostly balances out, but I wouldn’t be surprised if somebody has already run the numbers on that.
It sounds like what you’re saying is just that these are the situations where the old saying that a walk is “as good as a hit” applies, not that a walk is actually better than a hit. So it’s relatively more valuable than usual, not actually more valuable relative to a hit.
Nice work! I really enjoyed the probabilities that you brought into this. I’m a little confused about this statement:
“It’s tremendously unlikely, to the point of absurdity, that Kershaw had the same strike-throwing talent on 3-0 counts in 2019 that he did in his career until then. ”
I’m thinking that you should expect a pitcher’s 3-0 strike-throwing talent to be in-line with their career average. Why would this be so unlikely?
3-0 is situationally unique. Versus a weak batter and/or nobody on base, pitchers are more likely to stop aiming for corners and just dump a fastball in the middle of the zone. Versus a good hitter & RISP they might prefer to concede the base so they don’t give up a big inning getting their get-me-over pitch hammered into the seats (3-0 strikes when swung at get crushed relative to other counts). So I would expect some difference 3-0 in the strike rate and expect it to vary by pitcher, team, and circumstance.
That makes perfect sense. Thank you.
“While the league has only gotten more aggressive when it comes to flooding the zone on 3-0…”
This is probably because intentional walks no longer count in the 3-0 bucket. The chart shows the 3-0 zone rate jumping dramatically in 2017, which is when the IBB rule changed.
Yeah this looks right, and like something I should have caught. That explains the tick up; if you strip out 2016’s intentional balls, the zone rate on “competitive” 3-0 pitches was 62%.
Bonferroni correction, though. Did you in 2018 commit to “I’m gonna look at Kershaw’s 3-0 walk rate next year and publish an article whether or not it changes”? Nah, instead you were looking at stats and this one intrigued you, right?
It’s not easy to estimate how many non-intriguing stats you passed by in the process, but must be a few.
Very true. My process was to hunt for weirdness (more specifically, Meg asked me Thursday if I could write something quick for Friday and I said yes without any topic in mind). But now I can write another article based on observing the entire population of pitchers on 3-0 from year to year, which is great value. Good point!