Sunday, July 20, 2008

Clutch hitting: why do some studies find it, and some not?

A couple of weeks ago, I listed a bunch of clutch hitting studies that were unable to find any evidence of clutch hitting. They found that some players did indeed hit better in the clutch, but on a scale almost exactly that you'd expect by chance.

However, there were a few other studies that *did* find some clutch talent (albeit in very small quantities). Andy Dolphin found some here, and Tom Tango found some here.

Why the difference? I'm not sure; Dolphin and Tango didn't publish the details of their studies. But one difference is that the "could find any" studies were based on batting average and OPS, while Dolphin's study was based on on-base percentage.

So maybe the clutch effect is present in walks only? Walks are a big part of OBP, but a smaller part of OPS, and no part of BA at all. Maybe pitchers do more "pitching around" in clutch situations, and some players are better at laying off those pitches than others?

I don't know if that's the case, but it's one possibility. We'd probably be able to answer the question with more authority if one of these authors were to repeat exactly the same study, but using the metric used by one of the other authors.

Anyone have any other suggestions for why we might have these different results?

Labels: ,


At Sunday, July 20, 2008 3:45:00 PM, Anonymous Anonymous said...

I remember reading that pitchers tend to increase their BB rate in order to increase their K and decrease their HR rate with RISP, following the logic that an extra-base hit is more damaging than a walk in that situation. Tom Glavine's splits are notable for demonstrating this. Anyways, it seems more feasible that this would be a result of pitchers trying to paint the corners than some mysterious clutch hitting effect that manifests itself only in OBP.

At Sunday, July 20, 2008 4:22:00 PM, Blogger Phil Birnbaum said...

Hi, Alex,

I agree completely. But it can't just be pitchers, because the study adjusted for any league-wide change in OBP. So it has to be that different hitters have different OBP changes.

I think Andy found that the better sluggers had *less* of an OBP increase, which contradicts the obvious idea that sluggers get pitched around.

At Sunday, July 20, 2008 8:58:00 PM, Anonymous Anonymous said...

I think that some studies may find that clutch hitting exists simply because Type I error occurs a certain percentage of the time when statistics are analyzed correctly.

If you look for a p-value of .05 or less to reject the null hypothesis that clutch hitting does not exist, then 5% of the time, you will falsely conclude clutch hitting does exist.

At Monday, July 21, 2008 11:44:00 AM, Blogger Tangotiger said...

The entire reason is based on the number of trials for each sample. As I noted at some point, if you have a gazillion PA, your correlation will approach 1. In BBTN, Silver did find a correlation of r=.30 or .50 something. And, that is because the way he constructed his study, he got something like a few thousand PA for each player (he used the player's entire career). If you are going to do year-to-year comparisons, your number of trials is so low that in order to meet statistical significance, the gap is going to need to be very high, something that is very hard to do when your population of players are all so similar as hitters (most have an OBP of .300 to .400, which sounds big, but tell a statistician that all your samples are a true rate of mean of 34% give, with 1 SD = 3%, and see what he tells you).

The process I followed is exactly as noted here:

You figure out how much each sample is from the mean, figure the SD of that difference (z-scores), and compare that to what you expect from random (SD of z-Scores should be 1). If you don't get 1.00, then what you have is a bias somewhere, and the farther away from 1.00 it is, and the more players you have in your sample, then the more significant the finding is.

At Monday, July 21, 2008 12:00:00 PM, Blogger Phil Birnbaum said...


Isn't that exactly what Palmer and Ruane did? They looked at the actual distribution, and the theoretical (binomial distribution), and they were the same. But in your case, and Andy Dolphin's case, they were different.

So it's the same method ... the only difference is that you guys weighted walks more heavily.

At Monday, July 21, 2008 5:44:00 PM, Blogger Don Coffin said...

So here's an analogy to clutch hitting, where the analysis has fewer of the problems that plague the clutch hitting problem. (What are some of those problems? Well, for one, the hitter ain't in it alone. He's facing a pitcher who's also in a clutch situation, so the situation is the interaction to behaviors/talents/skills/etc.)


Look at golfers in "clutch" situations. Close to the lead, in the lead but only by a little. Do they play better, or do they play worse?

I know there's been some research done on this (but I can't find citations right now, which is always the way), and I think I remember the key finding.

Players who are closer to the lead following the cut tend to do better (relative to the field) than players who are further away from the lead. That is, they "elevate" their games.

Now, part of this is the reward structure in golf. Consider that the difference between winning the British Open and finishing second is that the winner's share of the purse is 18.37%, while the runner-up got 11.02%, and the players tied for 3/4 each got 6.24% (of a purse that was about $8.2 million). In short, there's a huge payoff to a small improvement in relative performance.

Hmmm. I think I'll look a little more closely at this...


Post a Comment

<< Home