The Evolving Link Between Strikeouts and Clutch
It’s really easy for me to get sucked into our Clutch leaderboards. We spend so much time focusing on true-talent levels, because true talent is what’s mostly stable, but timing is a huge part of winning or losing, and just because it’s hard to predict doesn’t mean it can’t be analyzed later on. We have win-probability numbers since 1974. On a per-plate-appearance basis, the most clutch position player has been Jim Leyritz. The least has been Ron Kittle. The most clutch active position player has been Willie Bloomquist or, if you figure he’s about to retire, Eric Hosmer. The least has been Giancarlo Stanton. Doesn’t mean Willie Bloomquist has quietly been a better hitter than Giancarlo Stanton. It’s just, timing closes the gap. (It remains an enormous gap.)
Because of the Royals, I’ve thought about Clutch a lot lately. Good offensive timing was a critical part of their run to the playoffs and then to the title. And when you think of the Royals lineup, you also immediately think about strikeouts, and about how they don’t collect very many. It’s only natural to try to establish a link between one and the other. I’ve tried to glance at this before, but now I’m coming at it in a different way. And I’m also trying to investigate a different but related point. Question No. 1: is there a link between Clutch and strikeouts? Question No. 2: if there is, has it changed at all lately, with velocity and strikeouts on the rise?
I didn’t have to do anything too complicated. I looked at players over five-year windows, setting a minimum of 2,000 plate appearances. To make things equal, I took that and calculated each player’s Clutch per 600 PA, representing something like a full season. At that point it was just a matter of looking at the Clutch/600 and the strikeout rates. You’re familiar with the basic statistical ways of trying to show a relationship. That’s what you’ll see in the table. There’s the five-year window examined, then the correlation coefficient between Clutch/600 and K%. After that, the slope of the line, with K% on the y-axis. The last column is simply the slope divided by 10, to show the change in expected Clutch/600 score per 10 percentage points of K%.
That paragraph was uninteresting to read, because that paragraph was uninteresting to write. This is the thing that’s interesting:
| Seasons | R | Slope | Slope/10 |
| 1976 – 1980 | -0.21 | -2.5 | -0.25 |
| 1981 – 1985 | -0.26 | -3.2 | -0.32 |
| 1986 – 1990 | -0.29 | -2.7 | -0.27 |
| 1991 – 1995 | -0.33 | -3.8 | -0.38 |
| 1996 – 2000 | -0.28 | -2.9 | -0.29 |
| 2001 – 2005 | -0.29 | -3.2 | -0.32 |
| 2006 – 2010 | -0.37 | -3.4 | -0.34 |
| 2011 – 2015 | -0.17 | -1.6 | -0.16 |
You might be more accustomed to seeing R-squared instead of just R. It tends to be R-squared that shows up on scatter plots, but R gets to show you direction. Granted, so does the slope. Anyway, here’s the first takeaway: there is a link between strikeout rate and clutch performance. It’s not too terribly strong, and you can go ahead and just mentally square those R figures to get a sense of what I mean, but this is picking something up. Generally speaking, as a player’s strikeout rate increases, his expected Clutch score decreases. The overall average is that, for each 10 points of K%, you get a change of about 0.3 Clutch/600, which is like three-tenths of a win. Small, but it kind of supports the Royals, who put a whole lineup of low-strikeout players together.
That helps establish one thing: contact ability can be good for your timing. Ever so slightly, low-strikeout hitters might be underrated, and high-strikeout hitters might be overrated. We’re talking little effects here, but we’re not talking about nothing. But then there’s the other question. How has the relationship changed, if at all? I’ve seen some people guess that contact might be more important these days, with so many pitchers throwing hard, and with strikeouts at an all-time high.
What the table shows is that, at least as Clutch goes, the relationship lately has been its weakest. Over the last five years, the link has been half as strong as it was in the five years before that, and the five years before that, and so on. You see eight progressive five-year windows. The link is fairly consistent, but it drops off in 2011 – 2015, with the second-weakest link being 1976 – 1980. A problem is I don’t actually know how to explain this. It might be simple, and I’m just being an idiot, but I prefer to think it’s complicated and in no way embarrassing that I don’t know what’s up. I’m not sure why strikeouts haven’t linked as well with Clutch lately. Maybe this is just randomness, and nothing to make anything of, but I have to report what the numbers say. Strikeouts and clutch performance: it appears that there’s something, but potentially less than there has been before. Huh.
As individual players in individual seasons go, I don’t think there’s much to worry about. The effect is small enough to be practically invisible, masked by other things. It gets more significant over several seasons, or when a team collects a number of low-strikeout or high-strikeout players. Then you’re adding effects on top of one another, and potentially talking about a win or two or three. The Royals had an extremely strikeout-averse lineup, so it makes sense they would have some Clutch benefit. But this shouldn’t explain all of it. Especially at a time when the link between Clutch and strikeouts is unusually weak. Even when we learn, sometimes we end with a shrug.
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.
Interesting.
I clicked on this title expecting the opposite conclusion, that a batter who is willing to swing for the fences and unfazed by the prospect of failure, would be mentally stronger in clutch situations.
Huh.
Not just contact, but speed as well. Beating out a hit with speed is sorta the “double layer” of protection against making an out. Kendrys Morales and I can both make contact but if he can’t beat out any IF hits or take 2B on a flare, he has a higher out probability than I.
Lorenzo Cain career OBP: .337
Kendrys Morales career OBP: .331
True, but OBP includes walks and HBP. Lorenzo Cain above is talking more about BABIP and baserunning ability.
Exactly right. Lorenzo at 1B is worth more than Morales at 1B. Little things like the ability to advance on a chop out = extending innings and run probability. Things like a flare down the line results in 1 guy at 2B and the other at 1B. Then there’s a GIDP. On and on. Cumulatively, a team of faster contact guys are worth > team of slower contact guys, even with same OBP
Maybe Lorenzo Cain should walk more since he’s so valuable on the basepaths.
He has a .337 career OBP. He doesn’t need to change a thing.
I made $3M last two seasons, and produced $92M of value. Maybe I should do whatever it was I was doing since I was probably more valuable relative to salary than anyone in the entire game.
My first thought about clutch decreases is that it probably relates to more emphasis on defense and defensive shifts. Better defenders are turning balls in play into outs more often. Hard contact is becoming more important for balls to be hit past defenders.
That was my first thought too, but for a different reason. With the shift on a lefty, the 1B holding a runner doesn’t really open up a hole anymore. Since runner-on-first is generally higher leverage than average, that could affect Clutch.
If the dropoff is mostly in lefties, that’s how I’d explain it, but if it’s righties too then I’m not so sure.
This is certainly an interesting thing to examine, but isn’t this result a bit expected? It doesn’t control for individual player’s talent levels. Lots of really really bad players strikeout a lot. As of right now, this just says that really bad players are worse in clutch situations. They’re worse in regular situations too though…
We can’t just look at correlations. We need some sort of panel regression
Clutch measures players against themselves, not their absolute performance in high leverage situations.
Although it’s much too small to prove anything one way or the other, Stanton’s league-lagging clutch stats are at least one data point showing the high-K/bad-clutch scores aren’t solely the result of bad players continuing to be bad. Would be interested in seeing that explored further.
I don’t think it’s actually true that “lots of really really bad players strike out a lot.” Unless you’re talking about NL pitchers trying to swing a bat. In fact, what really bad hitters do a lot of is make weak contact, and a whole lot of really GOOD hitters strike out a ton (hello, Kris Bryant, and many many others).
Not to sound mean, but we keep hearing from you guys that clutch players don’t exist in MLB. How can we reliably expect to link anything with clutch, given its flukiness?
I don’t think they usually say it doesn’t exist. They usually say it’s not predictive, which means nobody’s been able to prove it exists. This suggests that it doesn’t exist as a quality of certain players, but it doesn’t prove that.
And even if it isn’t a quality of particular players, it is still the case that clutch moments during games exist and sometimes players come through and sometimes they don’t. It’s possible to find large-scale correlations there. Descriptive, but not predictive.
One way or the other, it’s an interesting topic right now, what with the manner in which the Royals just won the World Series, so it’s cool they’re exploring it.
If you look in the glossary for CLUTCH, you’ll find some discussion, followed by a bunch of links.
One of which is:
Clutch skill DOES exist: http://www.insidethebook.com/ee/index.php/site/article/clutch_skill_does_exist/
This dates from 2009.
“Batters perform slightly differently when under pressure. About one in six players increases his inherent “OBP skill” by eight points or more in high-pressure situations; a comparable number of players decreases it by eight points or more.”
The reason clutch isn’t predictive is that it would take thousands of high leverage batting appearances (and also thousands of low leverage at bats) to determine who actually has it. The effect is real, and can be identified if you survey the league, but it’s simply too small to identify particular batters as “clutch” in any reasonable number of at bats. Random variation is too large and clutch too small.
why does that mean that batters perform better? couldn’t pitchers just be performing worse?
My stats are very rusty, but the R^2 for that last period is about .029. Isn’t that typically interpreted to mean that the variable being examined explains less than 3% of the variation in Clutch?
That’s exactly what it means. I don’t see any significance testing here either, so that effect might not just be small, it might not actually be bigger than zero other than by chance, either.
Well, between 2006-2010 R^2 was around .13, which, in these terms I think is significant. I think the question is not “does contact ability explain clutch” but rather “does contact ability have any effect on clutch” and it seems like yes it does.
I don’t think we can infer anything from this other than contact is slightly correlated with clutch.
I’ve been looking at the relationship between clutch and HR productivity. Just a handful of very good players on lineups that are either HR-heavy or rely on putting the ball in play. (By very good, I mean very good at their particular specialty).
I would really like to see you run a couple of regressions similar to what you provided here. The “clutchiness” of –
1)Each type of AB event, across all players, by year. Could be limited to just positive events.
2)Each team by AB event per year. Point is, do high HR teams with high Ks have a predisposition to lower clutch?
There are three types of events where I believe HR clutch could be disproportionately affected:
1) HR in blowouts. HR rate seems higher against mediocre pitching.
2) HR late in games where one run is all that’s needed.
3) The relationship of RP v Batter in late, close games when LI is high.
Item 3 is based on the clash of generally High K, low HR set-up RP and closers versus High K, High HR hitters. Seems to better fit high avg. hitters who make contact.
I am making no assertions here that the clutch v. HR relationship is significant. Still my small case studies indicate that these test runs might prove interesting.
What this approach would get at is not who is “clutch”, but what is the most clutch event, based on the difference between the two different WPA/LI ratios used to determine clutch for each player.
If common wisdom by many is true, there should be almost no variability between type of hit and clutch for the summation of all individual hitters. That is: clutch should be almost 0 for each type of event. (In request 1), listed above)
This basic question should be answered first. It deals with a common assumption that hasn’t been statistically verified.
Here’s a theory. Power hitters tend to strike out more, and contact hitters tend to hit the ball with less power. Shifting in recent years has diminished the value of weak contact, hurting the contact hitters more than the power hitters.
The follow-up question would be, have the Royals loaded up on spray hitters as well as contact hitters?
But power hitters tend to be the pull hitters, and get shifted a TON more. I think your last sentence is the more intriguing concept – spray hitters vs. pull hitters and their clutch scores.
Well, one thing that appears consistent is that BABIP is higher with runners on base than with the bases empty.
Looking at each of the last six seasons (2010 to 2015), non-pitcher BABIP with the bases empty is between .294 and .297 every year.
Looking at those same seasons, BABIP with runners on base is between .300 and .308 every year. (FWIW, 2015 was the highest total, at .308)
Looking at just RISP, it’s between .293 and .300. (FWIW, 2015 was the highest total, at .300).
So, combining what we see in the latter two splits, the highest BABIP each year looks to be with a runner on first.
What is the correlation between clutch and OBP? I would think players who put the ball in play regularly (or with a high BABIP) and/or take lots of walks would have good clutch scores.
Clutch is measured on the individual player level against the player’s own performance, not against the performance of players at large. Why would you except a low strikeout or high walk player to be better at hitting the ball or taking a walk in high leverage situations vs any other time?
Also, IIRC, low strikeout players are very different from high walk players. You have to get deep into counts to take a walk, increasing the likelihood of a strikeout.
I have a question, how does a batters strikeout rate affect how they preform against a strikeout pitcher? In other words, would your rather have a batter with a K rate of 15% and a wRC+ of 100 or a batter with a K rate of 30% and a wRC+ of 105 bat against a pitcher with a 10 K/9. I wouldn’t be surprised to see players that strikeout less be less affected by the strikeout rate of the opposing pitcher.
If that is the case it would make sense that it is easier to be more clutch if you strikeout less. Higher leverage situations occur at the end of games, when the bullpen pitchers are in the game. Bullpen pitchers, on average, have higher strikeout rates than starters.
So back to the example at the top.
You and Foster nailed one of the big issues with Clutch. Since better pitchers are being used in high leverage situations, hitting Clutch in aggregate will be negative – which has been the case every year this century (see link below).
And since the best relievers rely heavily on their strikeout ability, guys who don’t strike out as much have more immunity to this bias within Clutch.
http://www.fangraphs.com/leaders.aspx?pos=all&stats=bat&lg=all&qual=0&type=3&season=2015&month=0&season1=2000&ind=0&team=0,ss&rost=0&age=0&filter=&players=0
I absolutely agree with the first statement. And your second statement makes sense, and is what I would assume would happen, but I guess I’m curious to see if there is anyone who has studied the relationship.
Wouldn’t the main reason for low K% leading to Clutch be bullpens? The higher leverage situations tend to be the ones that have a team’s late-inning relievers, who average more K’s. The low-K hitters do comparatively better in those situations, helping their clutch score.
BigChief beat me to it
Isn’t it likely that in a clutch situation the pitcher also bears down and tries to make the best pitch he can? So any success by the batter is elevated by the fact he is facing the best a pitcher can throw. You can’t rely on just raw data. If a hitter hits .350 normally and hits .290 in clutch situations that is likely very good.
You said:
What the table shows is that, at least as Clutch goes, the relationship lately has been its weakest. Over the last five years, the link has been half as strong as it was in the five years before that, and the five years before that, and so on.
Just guessing, but might that be noise? The “true” effect of K% vs. Clutch-ness might be somewhere between the extremes that you’ve measured.
Yikes. Yes there seems to be something there, but that isn’t just small, its TINY. If you square those are R-values, and take the mean you get .07. And .3 wins per 10% K? Oh my. The standard deviation in K% in 2015 among qualified hitters was 5%. So you have to move 2 standard deviations to get .3 wins, but then the R^2 is just .1, so its easily over come by other factors.
I gotta say, it does look like there is something here, but its SOOO SMALL that in practical terms its basically nonexistent.
The estimated effect size for the Royals is about one win per year, which you could call small, or call it eight million bucks… even with that small r^2.
I wish we did 95% confidence intervals for everything we ever estimate, though.
Not surprising. To me, at least, it’s completely intuitive that three-true-outcome guys feast on the league’s shittiest pitchers and guys with more-contact-oriented approaches do better against the elite pitchers. Simply putting a ball into play against Clayton Kershaw is a big win for the hitter in a way it isn’t against Kyle Kendrick. Since better pitchers pitch more in high leverage situations (especially true in the playoffs) this tendency shows up even through the very noisy clutch data.
What I’m curious about is whether this is a symptom of strikeout rates or a symptom of a contact-oriented approach. Good pitchers also walk far fewer batters than bad pitchers. Not only do the Royals rarely strike out, they also walk less than every other team in MLB. Compare the Royals to, say, the 2015 Blue Jays who were a low K% team but also walked near the top of the league. Obviously with all the noise in clutch data, you can’t just compare a couple of teams, but I suspect K% is an inexact proxy for what’s going on compared to the percentage of PAs that end with the ball being put into play.
I’d like to see a statistical analysis of pitch locations, because as far as I can tell, the Royals were amazing at making contact on all kinds of pitches. I don’t know in how many of those cases it was poor pitch location vs clutch hitting – i.e. hitting a pitcher’s best pitch in that situation that went to the location that the pitcher intended.
I think there is going to be a re-evaluation of pure mph pitching as we go forward. The best hitters seem to be able to start timing them eventually in the game. This is just anecdotal but may account for at least part of the Royals’ late-inning rallies. Having movement as well as speed is more necessary than ever to try to fool these frighteningly good contact hitters.
A guess as to the reduced R value for 2011-2015. A combination of selection bias and smarter front office work. The selection bias is the 2000 PA cutoff, which means that in theory, lesser players get left off. Before 2010 those “lesser” players might have been judged more by RBIs and other things related to “clutch,” and thus not gotten the opportunity to amass 2000 PA. Now, we generally know better, and those players get the requisite PAs assuming they are actually good batters.
This is a pretty small effect, so the thing that’s most interesting to me is the change. The only thing I could think of here is the increased prevalence of shifts making the ability to just put the ball in play in a clutch situation less important. Basically, declines in BABIP.
Look at season-by-season numbers going back 10 years though, and there doesn’t appear to be any pattern of declining BABIP in high leverage situations – http://www.fangraphs.com/leaders.aspx?pos=np&stats=bat&lg=all&qual=0&type=1&season=2015&month=26&season1=2005&ind=0&team=0,ss&rost=0&age=0&filter=&players=0&sort=1,d
This relationship, particularly given that it is pretty weak/modest, makes a lot of sense to me. There’s often a feeling that, in certain types of the most important situations, high-strikeout guys can be pitched to in a way that better contact hitters can’t. I’m thinking about situations where the hitter really just needs to put the ball in play to change the game meaningfully, e.g. second and third, no outs, down by one run late (so the infield is probably playing back to avoid letting both runs score). There, the defense really needs a strikeout to get out of the jam. Lord knows this is what it felt like against the Royals; you just couldn’t expect to get out of those kinds of jams without allowing some damage.
Also, if high-strikeout guys tend to get more of their value from home runs, I would imagine that many of the most high-leverage situations are those where a single is all the offense needs, e.g. tying run at second or something. Because in the situations where you need a home run, the leverage is going to be lower since it’s so unlikely you get one (is my impression of how the leverage numbers end up working out).
Now, if this is the correct hypothesis, then I dunno how much it really explains the apparent shift in the past couple of years. Maybe everyone’s gotten worse at situational hitting, even the high-contact guys? I dunno.
I still want to see a study on the relationship between strikeouts and Team RE24-wRAA. It would be very interesting to see if context neutral stats don’t properly account for the benefits of stacking contact hitters.
I’m going to take a crack at that. I did take an earlier look at a base/out version of “Clutch” using RE24 and boLI from Baseball-Reference, and the Royals were above average but not near the top (the Twins and A’s had the highest “boClutch”, the Dodgers and Astros had the lowest).
The RE24 minus park-adjusted wRAA would be another good test though.
My theory is that what we’re seeing has to do with some macro trends in baseball strategy. In the 2000s, it became the rage to not worry so much about strikeouts, and both select for and train hitters to pull the ball and try to hit for as much power as possible even if that meant their strikeout rate would be higher. This worked at first, probably somewhat due to PEDs, but also because this was a successful strategy against the pitching and defensive deployment at the time. What’s happened since then, is teams started loading up on power pitchers and shifting their defense to counter the big-swinging bats (and are being helped by a growing strike zone, which is going to hurt a power hitter more than an aggressive, contact hitter). I would bet that some there are hard-throwing pitchers in the majors now who have success, but who would not have had success against more contact-minded hitters of previous eras, because their command isn’t actually that good. So the Royals, who’ve been deploying this strategy of going after contact hitters forever, are ahead of the times. Their hitters are guys who can neutralize the power, but relatively poor-control 2-pitch bullpen arms. If every MLB team hit like the Royals, there would be different types of pitchers in bullpens, better command, less power, more pitches to keep the hitters off balance.
I agree completely. A plausible strategy, I think, might be the rise of “two starter” games, where two pitchers who won’t go the whole 9 nevertheless have the 3 or 4 pitches, movement, and command of a starting pitcher to reduce the looks that contact hitters get in vs any one pitcher in the game. But relief will also need to adjust as the Royals continued their aggressive play in the late innings. Will be interesting to see if the Royals strategy continues to succeed or is countered. One thing I noticed is that contact hitters like the Royals seem to produce unpredictable / “ugly” batted balls, which I suspect is what makes them harder to defend.
I haven’t found anything about this, but does BABIP change significantly with runner on base? If it goes up, then low-K hitters are getting a boost with runners on, which should give them a bit of a boost in higher leverage situations as well.
There’s something on Tango’s site that shows the difference, but you’re absolutely right. It’s a significant enough difference, especially with a man on 1st vs. not, that your assertion is probably a big part of the correlation between contact hitting and Clutch.
I’ve spent a lot of time calculating WPA (which is used to calculate Clutch) by hand, and for anyone who’s done this the explanation for the correlation on is obvious. The great majority of High-Leverage situations have runners in scoring position. in this situation, if there are no outs and a man on 2nd, or 1 out and a man on 3rd, the batter is generally going to get a slightly positive score or neutral score for making an out but moving the runner along but an enormous negative score or striking out. In low leverage situations, even if there are runners in scoring position, the difference between a ‘productive out’ and a strikeout is much much less. Strikeouts just much more important than other outs in HL situations. When there are 2 outs, the situation is different, then it doesn’t matter what kind of out it is, but these tend to be a relatively smaller proportion of HL situations (for example, if a team is behind by two runs or three runs with a man on second, this will be much higher leverage if there are no outs or one out than if there are two outs.
A related difference–assuming that low K% hitters rely more on singles and less on HRs than do high K% hitters– is that in high-leverage situations the difference between singles and HRs is comparatively less than in low-leverage situations. If a team is behind by one run or the score is tied with two runners in scoring positions, either a single or a HR will get a very big positive WPA, obviously the HR will be higher but the difference won’t be so huge, especially if there are less than 2 outs. But if a team is for example behind by 2 runs with no one on base, the leverage will be lower, a single won’t have much WPA value, but unless it’s the last 2 innings or so a HR will have a much bigger relative effect.
Both of your posts are excellent points about why Clutch tends to be better for low K% hitters. That’s one reason I don’t like the name of the stat – it does measure a type of value/skill that the context-neutral stats might not capture, but it doesn’t necessarily have to do with the ability to perform better “in the clutch”
It seems like there is a missing correlation or at least a diminished weight given to real Clutch; which would be a hitters ability, as a percentage, to score a runner from second or third against that hitters ability to use all fields. The shift has hurt and exposed a lot of big bats, especially late in the postseason when everything is magnified. The Royals have proved that any contact against 98 mph fastballs puts infinitely more pressure on a defense as opposed to standing there while the Cubs lineup swings through fastballs.
Yes, I agree. Somehow pitchers have to make Royals-style hitters miss, or compensate on defense. Along those lines, I read a horrifying stat (for me as a Mets fan) that the Mets did not throw out a stealing runner at second after September 30. That has the probable effect of turning many walks and singles into doubles.
Another option is to try to counter with your own contact hitters.
Royals are 4.5% below league average on K%, which translates to +0.135 clutchiness wins. Over 8 batters it’s about +1 win above the context-neutral team WAR.
That’s not nothing. Especially given the low correlation — if more variance can be captured by adding some other variable, and the Royals are hitting on that, there’s a possibility (completely speculative) that this explains more than 1 win of their team Clutch.
Why the fascination with strikeouts and clutch? My point I suppose, is shouldn’t the best offensive players in terms of clutch be those that don’t strikeout as much AND don’t walk as much. In fact, I would think the best hitters would be below average walks and strikeouts, because most of their plate appearances would then end in contact. If this is true then that group would get the most plate appearances because they would play in more games on average (if healthy).
What about pitcher-batter match-ups? Since high K% batters are usually power hitters, when the match-up favors the batter, he’ll often be pitched around or walked. Low K%/low power batters don’t see that as often.