THE BOOK--Playing The Percentages In Baseball

Good points. Can anyone point out any personnel decisions that were ever made where a player’s clutch ability or lack of it influenced the decision? Trades or free agent signings.

My guess is that predicting a player’s overall wOBA using Marcel or some other projection will be alot more accurate than any prediction of how well a guy will do in the clutch.

I imagine there might be some psychological tests or something similar that you can do to predict this stuff. Unfortunately, we’ll never have access to what off-the-field info teams might extract from players. However, it’s not unreasonable to think that you can explain a significant portion of the variance in clutch game performance by some metric you extract in a testing room.
I say this with a degree of confidence since this is part of what I study with respect to brain-based performance measures. I don’t see why sport should be particularly different in principle.

mettle: you are ignoring empirical evidence.  The extent of clutch performance is 10 wOBA points per season.  At the extreme!

Do you know how hard it is to predict a change in 10 wOBA points based on PHYSICAL characteristics?  Now you are suggesting that you can predict a 10 wOBA gap based on EMOTIONAL and MENTAL characteristics?

Your best bet is if you can find the one blowout case, the Rick Ankiel case.  If you can find that one guy, then fine, I can believe that.  Otherwise, it’s virtually impossible to show based on whatever it is you think you can do.

The effect size may be tiny, but given the huge sample size, if an effect is there you can probably find it, if you had the data.
I’m not saying it’s a profitable or worthwhile endeavor from a team’s point of view—to squeeze out an extra 2 wOBA of predictive power—but I find the concept personally interesting. YMMV.

To me, the only interesting thing is if you can find the Ankiel case. 

Or, are Eckstein, Counsell, Willie Ballgame, and Juan Pierre really guys who can somehow age far better than others because of some intangible quality that manifests into tangible results?

That’s where the real interest is.  Looking to explain a 2 wOBA or 5 wOBA difference based on emotional or mental faculties?  That’s missing the really interesting stuff (Ankiel, Bloomquist) for the mundane.

MMDV (my mileage definitely varies)

How is the Fangraphs list determined? 

The list makes me think there is something other than clutch skill being measured here.  That’s very heavy on high-average hitters—Gwynn, Rose, Parker and Grace must have over 20 batting titles beween them.  Perhaps it’s a coincidence, but it looks like guys who are likely to deliver a hit (as opposed to BB or power) are being rated as more clutch.

The thing that always bothered me about “clutch” is that it implies that the player was holding something back during non-clutch situations.

If you could set up a league where salaries were determined by who was the clutchiest then *everyone* would be a clutch hitter - just by sandbagging during non-clutch at bats.

A guy like Tony Gwynn can afford to give away 10 pts of wOBA. He’d still be one of the best hitters ever. Conversely, his son needs to grind out every at bat if he wants to stay in the majors.

Guy:

sum(WPA)/aLI - sum(WPA/LI)

Tango:  I’m a little rusty on WPA/LI, but couldn’t that metric reward players whose skill set tends to be more valuable in high-leverage situations?  That is, players who produce a lot of hits?

As much as those of us who have read The Book, and a collective few outside of this population, “know” this, I still think there’s quite a bit of unintentional ignorance on this topic.

Guy, I don’t know.

To take an extreme viewpoint to illustrate your point: bases loaded, 2 outs, bottom 9th, tie game.  In that case, WPA/LI for any safe event will be +.054 wins, and the WPA/LI for any out event will be -.027 wins.

If say we have two guys who are equals:
.400 OBP
.400 SLG

.333 OBP
.513 SLG

To calculate WPA/LI in this case is easy enough:
+.054*.4 - .027*.6 = +.0054 wins
for the first guy.

The second guy is zero.

That’s for the right-half of the clutch equation.

In the left half, WPA for a safe event is +.333 wins and for an out it’s -.167.  So, the first guy’s WPA here is +.033 wins, and the second guy is, again, zero.

So we see that the first guy can build up WPA, and just take a small hit on his WPA/LI.

Excellent.  Ok, so it’s very possible that the REASON that we see clutch is not because of the skill aspect, but because of the biased nature of it’s easier to build up a clutch resume if you are a guy with a high OBP.

As I noted here:
http://www.hardballtimes.com/main/article/the-color-of-clutch/

One thing that was interesting is the kinds of playes Fans considered clutch. Overall, both our teams had a bit over 19,000 PA. Both had around 970 doubles and 80 triples. But my guys had almost 300 more homeruns, and 600 fewer singles. My guys had 500 more walks and 1000 more strikeouts. ... It’s very possible that to a fan, clutch is all about doing what Carlos Beltran didn’t do in his last at-bat against the Cards, when he took strike three. ... The Fans have a clear bias as to what they think is clutch: put the g-dd-mn bat on the g-dd-mn ball. This bias is best exemplified by Reds fans, as I noted before the season started:

The Reds Fans detest their best hitter (Adam Dunn) so much that they actually selected four different hitters ahead of him. Every time I would check the results, a new leader would emerge. Ken Griffey Jr., Scott Hatteberg and Brandon Phillips each would have made a fine choice, but the task will be taken up by Edwin Encarnacion. (And Javy Valentin was just behind Dunn in fan appreciation.)

#9/Guy: Yes. Also guys with large platoon splits, like Granderson. You’ll let him hit against a right-hander in the earlier innings and in blowouts, but in a game that’s close-and-late? Nah, he’ll see the LOOGY.

Then again Guy, maybe I spoke too soon.

http://www.baseball-reference.com/players/split.cgi?id=gwynnto01&year=Career&t=b#lever

Gwynn shows he did much better in higher leverage than low leverage.

Same thing with Vizquel:
http://www.baseball-reference.com/players/split.cgi?id=vizquom01&year=Career&t=b#lever

Dave Parker too:
http://www.baseball-reference.com/players/split.cgi?id=parkeda01&year=Career&t=b#lever

So, I take back (some, if not all) of what I said in my previous post.

Tango and Mettle,

Couldn’t this be similar to control of BABIP by pitchers?  That most of the variance in skill gets lost as a pitcher advances to higher levels?  A major league rookie may not have faced a quality cutter before, but he has batted in clutch situations hundreds of times since he was 10 years old.  Almost all of the good athletes who could not make it mentally don’t make it to the majors, greatly reducing the variance in the MLB population.

Yes, naturally.

Maybe I missed the discussion, but perhaps the wrong conclusion is being drawn.  Yes, the leverage splits confirm that some players hit better in high leverage situations, but from where I see it, I could just as well draw the conclusion that these players didn’t “step up their game in clutch situations”, but sluffed off in low leverage situations, inflating the appearance of a “clutch” ability.  Likewise I could argue that some of these players could be experiencing a “warm up” effect, where it takes them a few innings to get into a groove, and when they do, it’s in the later innings, which are generally of higher leverage value.

Again, if this was already discussed (and refuted), I apologize.

kds

I did one small study that showed that batters with experience do better in close and late situations. That is, they did not drop off quite as much as players with less experience.

http://cyrilmorong.com/Clutch-experience.htm

Cy

Using sum(WPA)/aLI - sum(WPA/LI)

as the measure of clutch, I wrote the following article

“Don’t Let Your Little Leaguers Grow Up To Be Right-Handed Power Hitters Who Strike Out Alot Because They Might Choke In the Clutch”

http://cybermetric.blogspot.com/2010/07/dont-let-your-little-leaguers-grow-up.html

Using regression analysis, being right-handed, striking out alot and hitting alot of HRs were all negatively and significantly related to this measure of clutch hitting.

Whenever these studies are done it seems to me that they are comparing what a hitter does in normal leverage situations to what they do in high leverage situations and call it clutch.  The basic conclusion is yes clutch exists but the amount of skill is minimal.

Has anyone looked at the issue the opposite way? In other words, does choking exist?  Do we find that certain types of batters consistency perform worse in high leverage situations over the course of their careers.

To me, maybe clutch shouldn’t necessarily be define as being more productive in high leverage situations but rather maintaining your normal production and not doing worse (ie. choking).

Cy, thanks for link.  One thing we’re measuring here is batter handedness:  every one of these “good clutch hitters” was a LHH or a switch hitter.  And we know that these hitters have an advantage with a runner on first (as will be the case in most high-LI PAs).  So you certainly want to adjust for that.

Tango, you show how high-OBP hitters may have an edge.  That makes sense, but I was suggesting a slightly different angle on it:  that high-BA hitters will be considered more clutch using this metric.  In your example (bases loaded) a walk has a lot of value.  But in many high-leverage situations, a walk isn’t that valuable relative to a hit.  And similarly, a HR often has less value than usual relative to a non-HR hit.  So my hypothesis is that this definition of clutch will tend to privilege high-BA hitters in general, and LHH high-BA hitters in particular.

I guess the way to figure this out is to measure the value of each offensive event in high-leverage situations, compared to its overall average value.  Hitters who tend to produce the events with the biggest differential will look clutch.  My guess is that singles and doubles have the biggest differential, along with outs.

Guy, I could have used a different illustration to keep walks down (1b open), but singles up, yes.

***

In The Book, Andy took did a better control, by only using opposing pitchers that are RH, and focused on innings/score, so that there would not be any kind of base/out bias.  And he just looked at overall wOBA in these two splits.

So, he addressed the issues people are looking at in this thread, and he did find a clutch skill.

I know Andy found more variance than random chance would produce.  And that probably means there is some clutch/choke skill (although there are other possible sources of variance).  But I think we need to make a distinction between two ideas of “clutch performance”:

1) hitters who performed better in the clutch than they did in non-clutch situations (for reasons other than the quality of pitcher faced)

2) hitters whose skill set is well matched to clutch situations. 

I think most fans mean definition #1.  And that’s what Andy was looking at.  But the WPA/LI metric is to a significant degree a measure of #2.  Looking at hitters with at least 2000 PA in the last decade (2001-2010), I get these correlations for Clutch/PA:
BA .11
BB/PA -.12
ISO -.29
OBP -.05

Clearly, BA is rewarded while BB and HR are penalized.  So OBP is not what matters (small negative correlation).  If we look at BA/OPS—a rough measure of the share of a hitter’s value derived from BA—the correlation with clutch is .31.  And this is all consistent with Cy’s findings.

All of which just reflects the reality of how those factors’ value changes in high-leverage situations.  And it’s consistent with what fans said they wanted in a clutch hitter.  But your “Clutch” statistic is telling us what kind of hitter someone is, as much as the player’s ability to adjust performance to circumstances. As such, I think it’s poorly named.

Admittedly, there’s not a bright line here.  The fact that LHHs have an advantage with a man on first is a real advantage in many clutch situations—so maybe that should “count” as clutch.  But being a low-walk, high-average, limited power hitter?  I don’t think that’s anyone’s idea of what “clutch” means.

*

Turns out we already hashed some of this out: http://www.insidethebook.com/ee/index.php/site/article/best_and_worst_clutch_hitters_of_the_retrosheet_era.  (Someday we’ll be like the old joke about prison inmates:  instead of making arguments, we’ll just say “thread #1257” and we’ll all know what that means.)

Guy: totally forgot about that thread.  Post 30 was my best stab at explaining WPA/LI (Situational wins).

Guy, you’re welcome.

Tom, I tell jokes in class and I had to start numbering them because the students told me I was repeating them. So one day I said “#3” and all I got was groans and a few minor, scattered chuckles.

Then one of the other teachers stuck his head in the door and said “#3” All of my student fell out of their chairs laughing hysterically.

I asked them why the laughed so much harder the second time when my colleague said “#3”. It was because he told the joke much better than I did.

Hopefully this is the correct one:

http://www.youtube.com/watch?v=mb615ASwvFc

@21: I ferget, being only 20-or-so copy since I was last reading through The Book (bathroom reading), but did Andy come up with a profile of a clutch hitter akin to Cyril’s? It would be nice to know the gap between expected and observed clutchiness (as opposed to truthiness clutchiness) for players before we can spout how so-and-so has shown such great strength of character in clutch situations.

I am wondering a few things about “clutch” when it comes to pitchers.

1.Since pitchers change their approach in different situations (going right after hitters in blowouts vs. being more careful in close games, etc.), it would seem that pitchers have more personal impact over “clutch” than batters would.
2.  If this were true, why do we see career batting leaders with much larger “clutch” numbers, positive and negative?  It looks like the top “clutch” batters are near +20, but the top career pitchers are around +10.
3.  I know I have seen a 5,000 PA r=.5 point for batters and “clutch” somewhere.  Has any research been done to determine when r=.5 for pitchers?
4.  If “clutch” is something that we can quantify for long-career pitchers, can we include it into WAR?  Since Smith’s WAR already “includes” situation pitching to some degree by using RA - do we need to weed out the runners-on-base situations in “clutch” before adding it to WAR if we chose to do so? 

Thank you, experts!

#7/Matt

You’ve nailed my thoughts on the subject. 

I’ve always thought “show me someone who can consistently step up their performance in the clutch and I’ll show you somebody who performs short of his abilities 80% of the time.”