A while back, our own Julian Levine noticed something odd when perusing the leaderboards at FanGraphs--historically, power hitters seemed to be worse when it came to their Clutch score.
I took a quick look at roughly 800+ players (≥3000 plate appearances for their career), and sure enough, the higher a player's Isolated Power (ISO) or Slugging (SLG), the lower their Clutch score. A player's ISO had a -.414 correlation to their Clutch score, while SLG came in at -.336.
Lots of people have taken on the issue of Clutch in terms of whether such a talent exists and whether we can measure it. (This 2009 piece by Tango is a pretty good place to start.) I don't want to get into a debate about the existence of Clutch at the moment, but rather try to understand why this relationship exists at all.
So let's start out with some general correlations of other hitting outcomes/attributes and Clutch:
Generally, the correlations are not very strong. We see the negative relationships between Clutch and things like wOBA, ISO, and SLG, but even ISO only explains a little over 17% of the variance in Clutch performance.
What's interesting is that the strongest positive correlation we see is for ground balls (GB%).
The higher your GB%, the more "Clutch" you turn out to be.
This jives with what we see regarding power (ISO) and overall hitter ability (wOBA). GB% has a negative relationship to wOBA (-.291) and, not surprisingly, ISO (-.527). Better hitters hit for more power (.792), and more powerful hitters tend to hit a higher percentage of fly balls (FB% correlated to ISO at .583) than ground balls (-.527).
Moreover, power hitters tend to strike out more often. K% is correlated to ISO at .611, and to Clutch at -.339.
So why are ground ball hitters seemingly more clutch?
One idea is that in high leverage situations defenses play hitters differently. For example, often times managers will draw in their infield with less than two outs. I don't have specific numbers, but ground balls already have a higher likelihood of turning into hits. Pulling in the infield should increase that likelihood, therefore boosting a ground ball hitter's chances of getting a hit in these situations.
So it may be that these hitters aren't blessed with the ability to "raise their game" in high leverage situations so much as the defense on the field is positioned in such a way that plays to their strengths.
Another possible explanation, suggested by Matt Klassenn, is that better hitters tend to be more powerful hitters (which is true--correlation between wOBA and ISO is .792, .899 with SLG), and those hitters have bigger platoon splits. Given that, these hitters are being matched up against same-handedness relievers in higher-leverage situations.
5 recs | 14 comments
The Shift
This is a fantastic article. I’ve said for a long time that if we want to understand clutchness, we need to look for players who are “structurally clutch” rather than players who have some Tebowsical mental power. We need to look at players who are stacked up to succeed more in clutch situations. Interestingly, I found that lefty power hitters who are shifted against are significantly better on balls in play than right power hitters in this article (free, per my request, due to the help of hundreds of SB Nation commenters who helped out on 30 sites):
http://www.baseballprospectus.com/article.php?articleid=11861
This was only true of BABIP clutchness, not HR,BB,K clutchness, so it seems like the limits in employing the shift with men on make hitters more clutch. I suspect there are lots of things like this, and the example you give about platoon splits is probably a big part of it.
Matt Swartz - January 11, 2012
Outstanding post. These data seem to support the conventional wisdom that you don’t want a power hitting high-strikeout guy up there in a clutch situation, although perhaps for different reasons than some fans think. Defensive positioning does seem to be a good explanation for the ground ball correlation.
LPanas - January 11, 2012
Holy Crap™, Bill!
Justin Bopp - January 11, 2012
How much of this might be due to team walking the better hitters and pitching to worse hitting in clutch situations?
Or even “artificially creating” clutch situation for these bad hitters by giving them more people on base after said intentional walking?
mattybobo - January 11, 2012
This
Rochestie4ever - January 13, 2012
Great work Bill
Also, love charts with black backgrounds.
J-Doug - January 11, 2012
This is awesome.
Fantastic stuff, Bill.
Julian Levine - January 11, 2012
Really good stuff. A couple thoughts.
1. Before I got to the end, I immediately thought: “Dangerous” hitters (more power, higher wOBA, etc.) are more likely than a banjo hitting shortstop to face a reliever brought in to get the platoon advantage in a key situation, and will accordingly perform worse than they will overall. I think this is huge.
2. (Follow me here): The highest LI situations are those late in the game with 2 outs, close or tied, men in scoring position. It is in these situations that a player’s Clutch factor is going to be made or broken.
(See http://www.insidethebook.com/li.shtml)
Two outs precludes the double-play, which makes groundballs relatively more valuable (less potentially unvaluable) than in the “average” situation.
Moreover and more importantly: in these situations a base hit is often as useful or almost as useful as a home run, which means the non-power hitter’s skill set is relatively more likely (vs. a HR) to jack up his WPA by a considerable margin than it would be in a neutral or less extremely clutch situation.
The flip side follows: a home run is LESS valuable (per WPA) relative to a single than it would be in neutral or less extremely clutch situation (in either of which it would still have a significant impact on WPA, but also on his WPA/LI), and accordingly a relatively greater portion of the power hitters’ positive WPA comes on those less clutch situation Home Runs.
Another way: A singles hitter can ONLY make huge WPA gains by hitting in the clutch, and the MOST extreme clutch situations penalize them LESS for being singles hitters than an average situation. OTOH A power hitter can make big WPA gains by hitting in any kind of reasonably close game, and the MOST extreme clutch (per LI table) situations penalize the WPA value of his home runs relative to the value of a single (as compared to their normal relative WPA values).
Think about a one run lead for the visitors, bottom of the 8th, nobody on. Because of the unlikelihood of a run being scored, the LI is only 1.3 in this situation. Duke Bigwhopper comes to the plate and hits a solo shot. HUGE WPA boost, right there! 0.34 or so, I believe (eyeballing the first tango we chart I could find). However, the WPA boost is NOT because of the overall leverage of the situation, but because he did the ONE thing he could do to dramatically swing the game’s likely outcome.
This means his WPA goes way up, but his WPA/LI goes up by almost as much – maybe even an equal or greater margin, because (a) there wasn’t much extra leverage to remove and (b) WPA/LI “knows” that he did the most relatively (vs. normal linear weights) valuable thing in that circumstance and valued that thing (again, relative to the other possible outcomes, not absolutely) accordingly, probably by at least 1.3 times as much as an average home run. (WPA/LI doesn’t JUST remove the magnitude caused by the leverage [the “de-leverage” part], it also realigns the neutral linear weights so the events are relatively weighted based on the game situation). (Calculating this is way over my head – I’m a conceptual guy.)
Would you call this bottom of the 8th, game-tying home run clutch? I would. But “Clutch” the metric does not really agree. Note that a singles hitter in this situation had no prospect of a big WPA gain.
So, you marry the fact that the “good” hitter is going to face a platoon disadvantage more often with the fact that a greater proportion of non-power hitters’ WPA gains will come in less clutch situations that accordingly lower his “clutch” rating with (polygamy!) the fact that a groundball hitter can’t create a double play with 2 outs and thus his 2 out [highest LI] performance will be relatively better per WPA) and there you go.
tobynotjason - January 11, 2012
Exactly
This is exactly what I meant by “Structurally Clutch.” Linear weights are shorthand. This exposes the gaps. I’d love to see a “Clutch/PA” regression a series of these variables to see.
Matt Swartz - January 12, 2012
Typo there, not sure exactly what you were saying in the last sentence.
But, yeah, there’s some circular logic to the metric since it finds the PAs where the game’s expected outcome historically changes the MOST depending on the outcome of that PA (which by definition need to capable of being affected by most players) and then awards points to the people who do better at changing the game positively (or, and this is the key to the GB thing, LESS POORLY at changing the game negatively) at those times than at other times: times when most any hitter is capable of swinging the outcome versus the times when it takes actual superb talent to swing it. The guy with suberb talent gets JOBBED for swinging the WE with his two out, bases empty home runs at most points in the game, even though (or rather BECAUSE) a home run is by far the best thing you can do with the bases empty and two outs, in the sense that it’s relatively MORE valuable vs. a base hit to RE/WE with two outs than with less than two outs/bases empty.
Take a look at this. Bottom 7, empty, 2 out, tie game: Leverage Index is 0.8. ANTI clutch situation, because USUALLY the game outcome doesn’t really hinge of the PA. Yet a home run swings the WE .243! A single is only worth .021 WPA (thus the LI of .8). The home run is super, super valuable and what is the effect on the hitter’s clutch rating? The slugger’s clutch rating will actually GO DOWN, because the (leverage 0) value of a home run that’s set at ELEVEN TIMES the value of a single (with all events/probabilities equalling 1) will be perversely more than the actual WPA gain.
Precisely because MOST hitters don’t do what he did, the situation was already defined as non-clutch, even though if we ask the question “when do you want to hit a solo home run, in terms of game score” the answer is “when you’re down by one or tied,” making what he did pretty much clutch-as-hell as far as most fans care.
My question now is: how big is this effect and does it overshadow what a bunch of seasons of clutch data can tell us? Should we just look at rate stats in “late/close” situations generically to avoid this goofiness?
This DOES make me wanna start paying attention to WPA/LI, though, since the actual game value of those 2 out home runs is so nicely captured there.
tobynotjason - January 13, 2012
the other aspect to consider is BABIP contributions; a GB-hitter will have a higher BABIP relative to a FB-hitter.
all other things being equal, a GB hitter is more likely to deliver that all important hit. while FB hitters may provide more net value, the value of an XBH relative to a single declines in high LI conditions, and the importance of BA rises.
tom s. - January 13, 2012
+1
My first thought was “High-leverage situations reward singles.”
WHYG Zane Smith - January 12, 2012
reply fail
+1 to tobynotjason, I meant.
WHYG Zane Smith - January 12, 2012
Lineup structure
Could this be another argument against traditional lineup structuring? For instance, batting a high contact/GB hitter lower in the lineup, in a spot where they may have more chances with runners on. Obviously the full spectrum of strengths and weaknesses in a hitter still need to be considered, but it seems like data like this good have an influence on how different types of hitters need to be valued, and perhaps the high GB types are being misused.
Excellent work BTW, enjoyed this very much.
VinBZ - January 12, 2012
You must Login with your SB Nation account and be a member of Beyond the Box Score to post a comment.