Clayton Kershaw Had Something Else Up His Sleeve
I’ve become atypically interested in Clayton Kershaw’s second arm slot. You know, the one where he drops from being so over the top. It is, apparently, Kershaw’s natural arm angle, but it’s not the one he took to the majors. It’s not something he ever used as a Dodger until he felt sufficiently inspired by teammate Rich Hill. Hill also drops down from time to time, and although Kershaw doesn’t drop down by so much, it’s interesting to see him messing around in the first place. Clayton Kershaw is, after all, the best starting pitcher in the world.
Players are always attempting some kind of tweak. They’re forever in search of some kind of leg up. Chris Taylor made the tweak he needed to make in order to become a quality major-league hitter. What interests me about someone like Kershaw is — a player like Taylor is strongly incentivized to improve. His career literally depended on it. Kershaw hasn’t needed to improve. Kershaw has only ever struggled relative to himself. Kershaw didn’t need to start changing up his arm angle. He wanted to try it anyway. Kershaw experimented for the sake of taking his opponent by surprise.
I love that drive that he has. It’s probably suggestive of how Kershaw got so good at all. He doesn’t want anyone to get too comfortable. To bring this all home: Kershaw has unveiled a couple surprises. Late last year, he suddenly started dropping down. And this year, one month ago, Kershaw threw a curveball. It was a special curveball.
This is Daren Willman. You know him for being the creator of Baseball Savant. Great site, great dude. Willman makes for a good follow on Twitter, and every now and then, I’ve seen him tweet out some variation of a Kershaw fun fact. Here is how Kershaw has used his curveball, over the course of his career, broken down by count.
The Kershaw curveball is a famous one. Vin Scully gave it a nickname all the way back in that one spring training. The curveball, in large part, is what got Kershaw to the bigs, and it’s a pitch he continues to trust. And yet, there’s something odd about the usage pattern. Kershaw throws his fastball in any count. He throws his slider in any count. He pretty much only ever throws his curveball when he’s even in the count or ahead.
It doesn’t seem optimal. You’d think that, when you’re a pitcher with three good pitches, you wouldn’t want to ever just completely ignore one. It’s a very uniquely Kershavian thing to do, but, because this is Clayton Kershaw, no one could ever give him too much grief. Obviously, whatever he’s doing has worked. In the big picture, Kershaw’s been the best. Can’t criticize the best for doing what he does, however he does it.
There’s no actual definition of being ahead in the count or behind. There are, in all, 12 possible counts. By the loosest definition, a pitcher would be behind in the count when there are more balls than strikes. To this, I’d like to add one condition: I don’t personally think a pitcher is behind when the count is full. I don’t think a pitcher can be behind when there are two strikes. So we’d be left with five counts that qualify: 3-and-0, 3-and-1, 2-and-0, 2-and-1, and 1-and-0. All of these counts are hitter-friendly. The numbers bear it out.
Using both Baseball Savant and Brooks Baseball, I did a lot of digging. When the count’s been 3-and-0, Kershaw has never thrown a curveball in his career. When the count’s been 3-and-1, Kershaw hasn’t thrown a curveball since 2008. When the count’s been 2-and-0, Kershaw hasn’t thrown a curveball since 2008. When the count’s been 2-and-1, Kershaw hasn’t thrown a curveball since 2009. And when the count’s been 1-and-0, Kershaw hadn’t thrown a curveball since April 11, 2011. He threw that curve to Pablo Sandoval. Then a whole bunch of nothing. Then, a curve. A curve in the year 2017.
There was, in 2012, this weird-ass blip.
That’s certainly different, but it’s not the classic Kershaw curve. Doesn’t count. Interesting! But, it doesn’t count. To put it all together, here’s a bar graph. Here are Kershaw’s total curveball counts when behind.
As the slider emerged, the curveball faded. Not out of the picture entirely, but out of the picture when Kershaw needed to battle back. Not a single behind-in-the-count curve in 2012. Not a single behind-in-the-count curve in 2013. Not a single one of them in 2014, or 2015, or 2016. And then, in 2017, one. And that one curveball occurred in the playoffs.
We’re going back to October 6. We’re going back to when there was still baseball, and the Dodgers were alive, and they were playing the Diamondbacks. The previous curveball in question was thrown in April of 2011. This one was thrown in October of 2017. That covered a span of well more than 200 Kershaw outings. It covered a span of well more than 4,000 Kershaw pitches when behind in the count. All of those pitches, and not one curve. And in the NLDS, Kershaw looked in at Yasmani Grandal, and Grandal didn’t ask him to throw a curve. Kershaw was in a 1-and-0 count, facing J.D. Martinez. Whatever Grandal initially called, Kershaw shook him off.
I don’t know how to explain it. Kershaw didn’t misread the signal. Kershaw and Grandal didn’t get crossed up. I spent annoyingly long staring at game video, and the call was clearly there, in plain view of the camera. After Kershaw shook his head a couple times, Grandal signaled for the curveball. Kershaw began his wind-up. I don’t know what it was. I don’t know why Kershaw wanted to go away from his pattern so suddenly. Maybe he just had a feeling. Maybe it was the confidence of pitching in a six-run game. Maybe, for some reason, Kershaw was just blissfully unaware of the circumstances. Kershaw, again, hadn’t thrown a curveball when behind in the count over some thousands of consecutive opportunities. For some reason or another, Kershaw decided to spin one in. It would be, objectively, one of the more surprising pitches imaginable. Martinez couldn’t have known it, but with him as the guinea pig, Kershaw was going to try to take advantage of more than five years of selective curveball avoidance.
Kershaw went into his wind-up. I don’t know how he was feeling. I’m not sure if he was feeling nervous, and I’m not sure if he was feeling excited. I’m not sure if he was feeling anything at all. Martinez was about to get ambushed. Maybe Kershaw just wanted to have a little of his own kind of fun. His arm went back. His arm went forward.
Baseball will never, ever, ever make sense. It is the worst thing about this dumb sport, and by far its greatest quality.
Jeff made Lookout Landing a thing, but he does not still write there about the Mariners. He does write here, sometimes about the Mariners, but usually not.
Amazing
Astounding.
Astonishing!
Flabbergasting…
shocking!
That’s incredible!
Inconceivable!
You meant (Inconceivable!)
Martinez is crazy. Between the time the ball left Kershaw’s hand and the time he hit it, you can very clearly see him go through a whole thought process:
Normal first reaction: “There’s the pitch.”
Surprise: “Whoa that’s a curve”
Acceptance: “Guess I better hit a dinger”
*hits dinger*
I had been wondering if this contributed to his career-high home run rate this season. Maybe he’s gotten (somewhat) predictable. I haven’t looked at the data on Baseball Savant, but if Kershaw has given up home runs when behind in the count at a higher rate than he did in previous years, it could show that hitters are sitting on his fastball or slider in favorable counts and crushing it.
batters have been able to “sit” on a fastball or slider when ahead in the count against Kershaw since 2010
looking at BR it definitely shows it….
1-0 12/36 2 2b 4 hr
2-0 7/15 2 2b 2 hr
3-0 0/0 bb
2-1 6/27 1 2b 2 hr
3-1 0/5 5 bb
25/83 5 2b 8 hr
was also 12/55 2 2b 1 3b 2 hr full count (1.5% of the time thrown)
and also 27/90 4 2b 5 hr 1st pitch(4.9% of the time thrown)
so 15 of his homers were thrown in counts where he threw the curve less than 5% of the time. In all the other counts- he gave up only 8 hr. .281 avg against in those non curve ball counts. .174 avg against when he wasn’t in those counts.
In playoffs- he gave up 8 homers
4 on 1st pitch of count
1 on 1-0 count
1 3-2 count
1 2-2 count
1 1-1 count
So of 31 homers total this year- he gave up 21 of them on counts where he would throw the curve less than 5% of the time.
he was also 17/55 with a 1-1 count- where Kershaw would throw the curve only 13.9% of the time. With 2 2b and 3 hr.
So in the 8 counts where he throws the curve less than 15% of the time- .286 avg with 18 hr. And then 7 postseason homers.
In the 4 counts where he throws the curve at least 15% of the time- .154 avg against with only 6 homers allowed. And 1 postseason homer.
1 last looking at it reply….
career numbers-
2017 18 hr in 15% or less situations 23 overall 78.3%
2016 3 hr in 15% or less situations 8 overall 37.5%
2015 11 hr in 15% or less situations 15 overall 73.3%
2014 4 hr in 15% or less situations 9 overall 44.4%
2013 9 hr in 15% or less situations 11 overall 81.8%
2012 11 hr in 15% or less situations 16 overall 68.8%
2011 10 hr in 15% or less situations 15 overall 66.7%
2010 9 hr in 15% or less situations 13 overall 69.2%
2009 6 hr in 15% or less situations 7 overall 57.7%
2008 9 hr in 15% or less situations 11 overall 81.8%
by count for his career:
23 hr 1st pitch
13 hr 1-0 count
6 hr 2-0 count
0 hr 3-0 count
15 hr 1-1 count
13 hr 2-1 count
6 hr 3-1 count
17 hr 3-2 count
93 hr in those counts out of 128 hr
But how does that compare with the rate that other pitchers give up HR? I would expect most pitchers are a lot more likely to give up long balls when behind in the count.
Just looking at Sale he only gave up 3 HR when he was ahead in the count(same definition as Kershaw). He gave up 11 hr when tied in count(0-0 or 1-1)
So ahead- Sale 3 hr Kershaw 8 hr Scherzer 8 hr Kluber 5 hr
tied 0-0 or 1-1 Sale 11 hr Kershaw 8 hr Scherzer 7 hr Kluber 7 hr
full count- Sale 2 hr Kershaw 2 2 hr Scherzer 1 hr Kluber 3 hr
looking at it as rates-
ahead-
Sale 3/107 Kluber 5/116 Kershaw 8/90 Scherzer 8/90
0-0 or 1-1 –
Kershaw 8/147 Sale 11/147 Kluber 6/138 Scherzer 7/121
full count-
Kluber 3/69 Kershaw 2/80 Sale 2/105 Scherzer 1/98
but then looking at Rick Porcello(led the league in hr given up)
ahead he only gives up 10/138 hr.
0-0 or 1-1 he gave up 11/184 hr.
3-2 he gave up 3/99 hr
gave up 9 homers when he was behind in the count. i think that’s where the good guys are so much stronger.
Baseball doesn’t just not make sense, it also laughs at clever plans.
looking at the stats, wouldn’t the Dodgers know this?
Hey Mr. Sullivan, you should have reiterated in your article that hitters swing something like 64.3% of the time when ahead in the count against Kershaw. You said it was nearly 4 standard deviations above League average. https://www.fangraphs.com/blogs/heres-an-astonishing-clayton-kershaw-statistic/
You’re correct! But they weren’t swinging at curveballs.
It’s one pitch and one at bat, but it is tempting to file this under “intuition vs. data”.
To be fair, Kershaw sort of telegraphed the pitch. He primarily throws three pitches, and when he shook his head twice, that should’ve been enough to tip JD off. At a 1-0 count, he can afford to look for a curveball and not swing if he doesn’t like it. He obviously liked it.
This was exactly my first thought. And in that specific count, he basically only throws fastballs/sliders. So, what else is there to shake off?
As a batter in an adult amateur league, this works very well at identifying pitches. If a catcher is shook off once and the signal was not a fastball, he will naturally signal fastball next. If he is shook off again, you can reasonably rule out fastball and look for something offspeed.
I hate to think that MLB is guilty of the same tendencies.
Great article. I feel like Kershaw’s curveball is his best pitch, yet he doesn’t use it as much as he should and relies too much on his slider. Should learn something about rich hills curveball usage
Sigh. OK Jeff, I’ll bite. I happen to believe Kershaw is the best pitcher in the live ball era. I guess that began in 1920. At the very least, there’s a great argument for that. EXCEPT in the post-season. I do NOT want to get into the why…it’s not relevant to my argument/point, which is…how can Kershaw be called “the best starting pitcher in the world” when, at precisely the moment he’s most needed to prove it, he’s failed so often? It’s not like he stinks; he obviously doesn’t. He’s been pretty good, mostly, but we’re not talking pretty good for the superlatives that Kershaw’s had applied. Haven’t we reached the point where the sample size is no longer considered too small to have meaning? And what of the 4-0 and 7-4 leads in game 5 this year? Do we just dismiss a failed key game (again) while labeling him the greatest?
I know…I’m gonna take my beatdown. Go ahead, I’ll pay the price on the downvotes just to try to have a conversation about this. Any beatdown that purports to excuse WHY he’s been merely good in the post-season, though, doesn’t mean anything. It doesn’t matter WHY.
I guess I’m one of the few that still find this topic compelling. Two questions that haven’t been explored enough:
1) If Kershaw’s postseason underperformance is largely a product of small sample size, why is there NO like-sized sample since his rookie year that is nearly as bad? Excepting his rookie year, his worst 19-start span in the regular year has a 3.66 RA/9; his 19 postseason starts show 4.76. Kershaw doesn’t even have bad months; his last 50 months all show an ERA under 4.10, yet his postseason ERA is 4.35.
2) If you dismiss ANY possibility that the pressure affects him, what do you say of his 6.59 RA/9 in 5 win-or-go-home starts? During his career, 16 others have made at least 3 such starts; Kershaw’s runs rate is the worst.
I’ve got pages of this stuff, but I think most folks just want to tune it out.
Thank you for saying in fewer words, and with analytics, what I was aiming for. “Tune it out?” There are fewer and fewer good baseball arguments to have anymore, especially with the brainy guys on this site. Anyone tuning it out just isn’t considering the question. For some reason, Kershaw is not held to the standard of other players. He’s given a pass.
There’s a very good reason that Kershaw is treated differently, because he’s obviously the best regular-season pitcher in baseball, no argument (although ironically the first time I really outraged Fangraphs’ readers was when I suggested early last year in the middle of Cubs mania that it was ridiculous to compare Jake Arrieta to Kershaw). I don’t think that the reason that Fangraphs’ writers and readers don’t want to realistically talk about Kershaw’s postseason problems is that they’re giving Kershaw a pass, and it has nothing to do with Kershaw in particular. It’s because they don’t believe that there is such a thing as players performing better or worse in the clutch. If there is evidence that popular beliefs about this for certain players are nonsense, Fangraphs’ readers and writers will point it out. The problem in Kershaw’s case is that there is just too much postseason data without an explanation, so that all that Fangraphs’ writers can do is keep silent and all that Fangraphs’ readers can do is give downvotes.
Ah, I see what point you are making. You were originally saying that the claim that Kershaw is the best starting pitcher in the world is open to question, because of his postseason record, for which he is being given a pass. I didn’t connect this post with the one above. Any claim that someone is the best requires some criteria for determining what qualities are being valued, and if all MLB performance is being given equal value statistically (which is how Fangraphs’ thinking works), Kershaw is clearly the best, because his mediocre postseason performance is only a small proportion of his total MLB performance. You would presumably claim that postseason performance should be valued much higher, to which a typical Fangraphs’ reader would respond that this is a small sample size which shouldn’t be valued more highly than any other MLB performance. There isn’t any way to resolve this argument because this depends on how different people define ‘best’ (I teach and do research in semantics, and a big point in this discipline is that different subcommunities can use the same words with somewhat different meanings, this is how human language works), but you will definitely not be able to convince very many Fangraphs’ readers to adopt your definition.
FWIW I personally would not trade Bumgarner even up for Kershaw, I really wouldn’t, and it isn’t a matter of Giants’ pride. For what I want for my team, the importance of winning championships overrides general quality, and I would much much rather have Bumgarner on the mound (and at the plate, for that matter) in a postseason game than Kershaw. Kershaw will give you a better chance to get to the postseason, but for me at least that isn’t as important. It’s possible that a few more years like 2017 might cause me to change my mind, it may be that I’d be happy just to see the Giants make the playoffs, but that definitely isn’t how I feel now. Does this mean that I think that Bumgarner is ‘better’ than Kershaw? I suppose in some sense it does, if you would rather have X than Y, it suggests that you think that X is better than Y. But I definitely wouldn’t seriously say (or think) ‘Bumgarner is better than Kershaw’ or even ‘Bumgarner is as good as Kershaw’.
Yes, sample size, clutch performance. I get it and believe it. Hell, it doesn’t matter if I believe it or not, because it’s a FACT. Which is why, each year, I still expect Kershaw to perform as well as he did all season. I can set my watch by my expectations. And each year, he disappoints. So here’s the question-would the typical Fangraphs reader EVER acknowledge that there’s an aberration that cannot be attributed to mere sample size when the sample size is no longer small?
For the sake of argument, I’ll throw out Don Newcombe. He pitched in 3 World Series for the great Brooklyn Dodgers teams, all versus the Yankees. During those 3 seasons he put up 13 total WAR. He was 64-20 during the regular season in these 3 seasons. He was the unquestioned ace of those staffs. And he was shelled in each series. 5 starts, 0-4, 8.59 ERA. He was so bad he even was benched by Alston! This is the epitome of a small sample size….right? My analytical side and my result side are at war over this question, no pun intended. His reputation was that he couldn’t win the big one. Is that unfair? And if so, at what point does that narrative (there’s that word again) change? Or does it ever change?
Looking at Newcombe- he’s the definition I think of SSS. I mean- he only had 22 postseason innings. To put this in context, Yu Darvish has already 26.1 innings in postseason.
This is where Kershaw is different- with over 120 innnings of work in the postseason(or over 5.5 times as many as Newcombe).
No, the typical fangraphs reader will not admit that kershaw simply plays worse in more important situations, no matter large the sample size is. It is against their religion.
1 thing that is really weird with Kershaw is in the regular season he’s only had 7 times where he’s started against the same team 2x in a row(that’s amazing when you think about it) .
In those 7 times-
1st start 50.2 ip 14 er 11/50 2.49 ERA
2nd start 49 ip 20 er 6/57 3.67 ERA
Could it really be that his stuff plays much better when you aren’t used to seeing it multiple times in a row?
it is amazing. Wonder what post-season back-to-backs against the same team look like, compared to regular season?
I wonder if part of it is batters getting more familiar with Kershaw the 2nd time seeing him in a series…..
he’s had 8 series from what I can see that he’s made 2 starts in a single series..
1st start- 18 er in 50.1 ip 66/11 k/bb ratio. 3.22 ERA
2nd start- 27 er in 46.1 ip 56/11 k/bb ratio. 5.24 ERA
I’ve thought about that possibility also but then I remember the cataclysm of the first game of the 2014 NLDS. On the other hand (I just realized after I read your post), if we dismiss that as a one-time fluke and take that game out, his 1st start ERA goes down to the low 2s, his normal range. Hmmm…
yeah how about 2.06. It makes more sense than a lot of the theories that we’ve heard.
And I would say that 2014 LDS start vs the Cards- he had started vs them the prior year.
1st start in 2013- 0 er in 6 ip 1 bb/5 k
other 3 starts vs them- 18 er in 16.2 ip 4 bb/24 k (9.72 ERA)
Cubs he’s had 4 starts last 2 years
1st start in 2016- 0 er in 7 ip 1 bb/6 k
other 3 starts vs them- 7 er in 16 ip 2 bb/13 k (3.94 ERA)
so it does seem like there is something with that where the more used you are seeing Kershaw the better you do.
This may be a normal thing though. I mean looked at Kyle Hendricks. He’s had 3 series last 2 years made 2 starts in same series
game 1- 1 er in 16.2 ip 9/18 bb/k ratio 0.54 ERA
game 2- 5 er in 16.0 ip 2/15 bb/k ratio 2.81 ERA
also he only started vs LA once this year- but that’s 3rd time in 2 years. He gave up 3 er in 5 ip in that start.
So it may just be that Kershaw has had the misfortune of seeing teams multiple times moreso than some others..
Looking at 2 Astros from this year:
Keuchel
1st 3 er in 13.2 ip 2 bb 13 k
2nd 7 er in 8.1 ip 3 bb 12 k
Verlander
1st 4 er in 15 ip 3 bb 18 k
2nd 2 er in 13 ip 1 bb 17 k
That’s pretty convincing stuff, John.
Most folks here want to tune it out because it goes against the orthodox Fangraphs’ ideology. This ideology is great in terms of enabling people to see through a lot of popular beliefs which have no basis, but in some cases it blinds (or deafens) them to facts which are staring them in the face (or are shouting in their ears). I have pages of stuff on how the Giants managed to win 3 World Series in 5 years with teams that didn’t appear to be very good, but they’re too complicated for everyday Giants’ fans to understand and they go against Fangraphs’ orthodoxy so there’s no point in bringing them up here. Some day there’ll be a paradigm shift and things like this will be seriously discussed, but I don’t know when that will happen.
I’m a Giants fan, too. I’ve attributed the 3 titles (well at least the latter two) as just getting into the playoffs, then riding good fortune and/or Mad Bum. And I’m good with that, proud even. To me, there’s great shame in a team consistently having great talent but not being able to close the deal (see Dodgers, Los Angeles). I suppose that’s my sample size issue coming through again. My obvious flaw is that I don’t have the analytical background or understanding to back this up. I am on FanGraphs (love it!) and obviously know I’ve got a LOT to learn, but I’m also old school. Anyway, I attribute winning World Series in this era as being good enough to get into the post-season, then being good AND lucky.
One thing-I’d rather have Kershaw. He is SO much more valuable than Bumgarner in helping a team get to the playoffs. But ideally, we’d smash them into one. Mad Bum Kershaw!
I was in a very very long discussion here about this a few days ago on another post. In fact I’m not sure it isn’t still going on. I don’t really have anything else to say about this here, except that I wish one or more of the Fangraphs’ writers would take a serious look at this, and I mean serious, not just waving it off as small sample size or cherry-picking stats to make it look as everything is really normal.
On a tangent, I nominate Jeff for the Hall of Fame of Declarative-Sentence Leads:
— “I’ve become atypically interested in Clayton Kershaw‘s second arm slot.”
— “I have learned something bad about the Royals.”
— “The biggest at-bat of Game 6 was when Chris Taylor doubled.”
— “No one ever wants to have to talk about the strike zone.”
— “I don’t know exactly what it is we just watched.”
— “The stat wars are over, and the bodies are buried.”
— “Being expressive is half of Yasiel Puig‘s whole thing.”
Jeff, I don’t know if you’ve read The Elements of Style, but you’re an avatar of its leading principle: “Eliminate needless words.”
You’re a great writer. Don’t ever change!