What If Clayton Kershaw Weren’t So Predictable?
Clayton Kershaw is the best pitcher on the planet.
You’re probably aware of his ability, and his hardware, which includes three Cy Young Awards, six straight top-five Cy Young finishes and an NL MVP trophy. But here’s the thing: he could be better.
Consider this fascinating nugget unearthed by Daren Willman:
Clayton Kershaw never threw a curveball when he was behind in the count last season… That's pretty crazy pic.twitter.com/vYSotPuzir
— Daren Willman (@darenw) February 27, 2017
Kershaw, as Willman notes, never threw a curveball when behind in the count last season. And that’s not all: as Jeff Zimmerman discovered in a December post at RotoGraphs, Kershaw has also been reluctant to employ his other breaking pitch, the slider, in hitter’s counts. A visualization made by Zimmerman of Kershaw’s pitch mix by count reveals the difference between it and the balanced approach utilized by Johnny Cueto. Cueto is willing to throw almost any pitch in any count. Kershaw, on the other hand, becomes extremely reliant on his fastball when he falls behind.
Overall, Kerhsaw threw curveballs at a 15.6% rate last season and at 13.2% rate for his career. It’s his third pitch, but it is also his greatest velocity-separation offering. He rarely throws a changeup. But while Kershaw rarely throws his slider in hitter’s count, he never throws his curveball.
And it’s not just fluky, one-year, phenomenon. Here are the total number of curveballs Kershaw has thrown in his career when facing a 1-0 count…
And a 2-0 count…
And a 3-0 count…
Nothing there! For his career, Kershaw has thrown 243 pitches on 3-0 counts. He’s thrown 242 fastballs and one slider.
For good measure, here’s his total curveball usage when facing a 2-1 count…
And when facing a 3-1 count…
It’s not as if Kershaw shouldn’t lean on his fastball. Since 2008, it’s produced the most linear-weight runs among all major-league fastballs (191.2) followed by Cliff Lee (141.1), Kenley Jansen (106.9) and David Price (101.6). But he also has the game’s second-best slider by that measure since 2008 (112.9) and fourth-best curveball (57.4). So it’s not if Kershaw lacks quality secondary pitches. We know these are excellent pitches. Said Vin Scully of Kershaw’s curveball “holy mackerel,” dubbing it “Public Enemy No. 1.”
Kershaw’s former personal catcher, A.J. Ellis, characterized the left-hander’s curveball as “unique in all of baseball.”
In 2016, Kershaw had the game’s fifth-ranking fastball (21.5), third-best slider (24) and 12th-best curveball (5.5) among starters who threw at least 100 innings. But when he gets behind, Kershaw trusts his fastball, and he trusts the pitch like few other arms in the game. Still, as great as Kershaw is, it stands to reason he could become even more dominant if he became less predictable.
There have been some attempts, such as this FiveThirtyEight piece by Neil Paine, to explore game theory as it relates to optimal pitch mix. Write Paine in that piece:
For the entire population of pitchers I looked at, a one-standard-deviation increase in optimality results in an earned run average about a quarter of a standard deviation lower. It’s not a huge effect (it only amounts to a handful of runs per season), but it does suggest that a pitcher can reap some benefit from trying to find an equilibrium that evens out every pitch’s effectiveness.
And an optimal pitch mix, for Kershaw, would of course be less fastball-oriented in hitters’s counts.
Consider: in 3-0 counts over the course of Kershaw’s career, opponents are slugging .800 with a .600 batting average versus the 242 fastballs he’s thrown in those 243 counts. You would assume that, if batters know what pitch is coming with 99% certainty, they would have better odds against it — even against a pitch as effective as Kershaw’s fastball. Batters can isolate not only zones, but also pitch type, in favorable counts.
In 3-1 counts for his career, Kershaw has thrown his fastball 462 times out of 526 total pitches, according to Brooks Baseball. Opponents have hit .308 with a .546 slugging mark against his fastball in those situation.
In 2-1 counts, Kershaw remains fastball dominant and opponents have hit .345 with a .525 slugging mark against the pitch.
Of course, there are dramatic swings in offensive performance when hitters are ahead in the count compared to behind. MLB hitters slashed .341/.342/.579 in 2-1 counts last season, in line with Kershaw’s career numbers with his fastball in such counts. But this is Kershaw. He typically isn’t near league average in anything.
Consider tOPS+, via Baseball Reference, a measure that quantifies how well a player performed in a specific split compared to his own overall performance:
| Batter ahead | Even count | Pitcher ahead | |
|---|---|---|---|
| League (2016) | 173 | 92 | 40 |
| Kershaw (career) | 198 | 102 | 22 |
| Kershaw (2016) | 181 | 140 | 22 |
Now, Kershaw doesn’t fall behind often. He walked less than a batter per nine innings last season. On 425 occasions last season, Kershaw faced a count that was even or in his favor. Only 119 times did he face a situation where the batter was ahead in the count. MLB hitters still posted an anemic .657 OPS when they were ahead in the count, but when Kershaw was ahead it was an astounding .290 mark. MLB hitters, overall, posted a .998 OPS when ahead in the count last season and a .521 mark when behind.
So as great as Kershaw is, perhaps with an added element of unpredictability he could be even better.
A Cleveland native, FanGraphs writer Travis Sawchik is the author of the New York Times bestselling book, Big Data Baseball. He also contributes to The Athletic Cleveland, and has written for the Pittsburgh Tribune-Review, among other outlets. Follow him on Twitter @Travis_Sawchik.
Honest question here, since this is the first time I’m hearing about tOPS+: wouldn’t this metric have a tendency to be skewed for pitchers who are ahead/behind in the count more often than average?
The fact that his tOPS+ when behind in the count is worse than league average seems to be a natural outcome of Kershaw’s tendency to work ahead in the count more frequently than the average pitcher – thus skewing the data set that makes up his “own overall performance.”
On the flip side, I would expect pitchers with control issues to have lower tOPS+ scores when working behind in the count, because being behind in the count is relatively closer to their natural state.
It’s hard to get at the “batter ahead tOPS+” data by pitcher on the B-R website, but looking at it by team seems to support my suspicions that situational tOPS+ can be skewed. In 2016 the two teams with the highest/worst tOPS+ in “batter ahead” situations were the Cubs and Dodgers. The two best were the Rockies and Twins…
Really weird thing with the video above – it takes the audio from Vin’s famous Spring Training call of Kershaw in 2008 that I preserved and puts it with an entirely different clip from a future year. Here’s the original clip. http://dodgerthoughts.baseballtoaster.com/archives/920274.html
video removed for copyright 🙁
I wonder if part of the reason is because curveballs in general (& I’d imagine Kershaw’s in particular since it moves so much) have a lower called strike percentage than other pitches?
Absolutely. But never? Not once in eight years just to give a different look? That’s astounding 🙂
Kershaw has been quoted giving this as a reason for why he has started throwing sliders more frequently instead of curveballs.
Is that last table backwards? You said Kershaw doesn’t fall behind often, but it says batter ahead: 181 and pitcher ahead 22
That’s the batter’s OPS against Kershaw in that particular situation, relative to Kershaw’s OPS against in all counts combined. It’s not a counting stat.
It’s a rate stat, where higher is bad and lower is good. Kershaw doesn’t usually fall behind against hitters, but when he does, batters do better against him than they do against a league average pitcher.
I think tOPS+ is actually not against league average pitcher but against average Kershaw. So batters who are ahead of Kershaw are a LOT better than batters in the average situation against Kershaw. The point of that last table is that Kershaw’s ahead/even/behind split is even more extreme, relative to himself, than the league as a whole is to itself, although of course bad Kershaw is still better than the league-wide average.
Actually, what that table tells me is that he has approximately the same relative split to himself that the league does, begging the question whether he actually could get any better this way?
I’m a huge fan of the data visualization from Zimmerman on the pitch mix. Would love to see it on player pages.
Um… “For his career, Kershaw has thrown 243 pitches on 3-0 counts. He’s thrown 242 fastballs and one slider.” Article on that one slider, please!
Talk about burying the lede. “That one time Kershaw didn’t throw a fastball in a 3-0 count”
Good odds it’s a pitch f/x malfunction.
That slider probably got 3 strikes at once and the catcher threw out both runners trying to steal.
I’m hopping on my golf cart next, and heading over the the Ranch and tell Honeycutt about this. That’ll fix those pesky hitters.
I think we’re splitting hairs at this point. I would guess that more harm than good would come from Kershaw forcing more curveballs than he naturally feels comfortable throwing.
Maybe eventually he’d get more comfortable with the new pitch mix, but if I were his pitching coach, I’d be inclined to leave Kershaw to be Kershaw. Unless he came to me with the suggestion, that is.
Loved the article! Watching that video makes me excited for the season, but also reminded me that we have to watch it without Vinny
So if a hitter takes a pitch In an “ahead” count is that reflected in this tOPS+? Basically I’m asking how often did batters put balls in play agaisnt Kershaw in ahead counts? If he’s locating and getting strike calls or swings and misses that turn the at bat more in his favor that wouldn’t be considered here right? maybe the batters wouldn’t hit the curve or slider as well, but it’s also very possible that this would get Kershaw deeper into bad counts. I think it might be better to approach this as “what can we learn from Kershaw” (because he’s amazing), not “how can we fix him”
Obviously small-sample-size caveats (and SSS may still be most of the explanation), but is it possible his postseason struggles could be in part due to this predictability? I’m more of an NBA fan, and people often talk about teams preparing more for specific opponents in the playoffs in the NBA, so this could be a possibility in the MLB as well. Kershaw’s career xFIP in the postseason is similar to his career xFIP in the regular season, but both his FIP and (especially) ERA are much worse. This suggests he doesn’t struggle much in terms of BB and K’s but is in terms of BABIP and HR/FB. Both of those obviously have high variance in such small samples, but could also in part be a result of simply better contact being made. His LD% is similar in the postseason vs regular season, but his FB% is higher (and, obviously, his GB% is lower). Would be interested in seeing more detailed batted-ball data (velocity?) between his postseason vs regular season…
Shocking stats in here. Clayton Kershaw has faced 243 3-0 counts? When did any of those happen?
Also, just looking at that graph makes the idea of standing in against Johnny Cueto very uncomfortable. At least against Kershaw, you know it’s a fastball or slider that will be getting you out.
I had no idea Kershaw was that predictable in certain counts. To think he still has room for improvement!
Perhaps Kershaw is predictable, but it’s also safe to bet that Kershaw’s walk totals will be nearly non-existent. If this is how he keeps that so, by all means not the worst thing I would think.
“Consider: in 3-0 counts over the course of Kershaw’s career, opponents are slugging .800 with a .600 batting average versus the 242 fastballs he’s thrown in those 243 counts.”
I think some context is needed here. According to B-R (http://www.baseball-reference.com/leagues/split.cgi?t=b&lg=MLB&year=2016), in 2016 the MLB hit .402 with an .820 SLG in 3-0 counts. Suddenly, Kershaw’s .600 and .800 don’t seem all that horrendous.
“But this is Kershaw. He typically isn’t near league average in anything.”
It doesn’t necessarily surprise me that much that Kershaw has close to the league-average batting-against line on plate appearances that *end* on 3-0. What I want to know if he prevents those outcomes more often. If batters take more of his 3-0 fastballs, or if they put them in play less often, then that would count towards him being better in those counts, even if he is about average when the ball is actually put in play.
When you cite a .600 AVG and .800 SLG giving up by Kershaw in 3-0 counts over his career, do you means ON a 3-0 count or AFTER a 3-0 count? Baseball Reference lists two separate statistics – “3-0 Count” and “After 3-0”. The league, as one would intuit, has much better numbers ON a 3-0 count than after a 3-0 count.
I think it should go without saying that Kershaw should throw way more fastballs when behind than when ahead. Even though it should go without saying, I will say the reasons:
-He has better command of his fastball than his other pitches, meaning he is more likely to throw the pitch in the zone
-The fastball is more likely to be called for a strike when in the zone
-Batters are less likely to swing, and therefore more likely to take a strike in the zone
-Kershaw is generally hard to hit, meaning batters are looking to take a walk more than they would for another pitcher. You would expect batters to be even less inclined to swing at a Kershaw pitch up 3-0 or 3-1, for instance.
-Kershaw’s fastball is hard to hit, which counters some of the benefit the batter gets from its predictability.
-His slider and curveball get much of their effectiveness from a high swing rate and low zone rate. Considering batters are less likely to swing, it particularly helps his fastball, which he often uses to get calls in the zone, and it particularly hurts his breaking balls, which tend to get swings out of the zone.
In other words, there are many reasons why his fastball is a better choice and other reasons his slider is a worse choice. I think these reasons are magnified for Kershaw because I would guess he shows higher swing splits by count than the typical pitcher, and his pitches are more distinct from one another than for the typical pitcher. Like I said above, his relative-to-himself splits don’t even look that out of line to me, especially considering they are ratio stats, and his OPS-against is so low that you might expect his relative-to-himself ratios to appears larger due to a smaller denominator.
I think tOPS+ badly distorts the picture of Kershaw’s ahead-or-behind effectiveness relative to other pitchers.
Fact is, Kershaw last year allowed the lowest OPS among regular SPs in *both* splits. And both were 2nd-best of the last 5 years, of those with 20+ starts.
tOPS+ distorts Kershaw because his overall OPS is so low, even small raw gains by batters when they’re ahead register as big percentage gains. But if you compare in raw OPS, you could argue that Kershaw actually *gains* effectiveness when behind, relative to others. Last year:
– When ahead, his OPS allowed was .231 below MLB (.290-.521).
– But when behind, his OPS allowed was .341 below MLB (.657-.998)
This is the most shocking Fangraphs article I have read in awhile. The predictability… Are other pitchers like this, or is it just Kershaw?
Okay. I got curious about Kershaw’s number on 3-0 count.
http://www.baseball-reference.com/players/split.fcgi?id=kershcl01&year=Career&t=p
That .600 AVG and .800 SLG number on 3-0 count
comes from batters having 6 hits out of 9 AB including 2 2B’s.
So…. Super Small Sample Size anyone?