It seems like this week(end) everyone has been talking about clutch hitting. It's mostly in response to JC Bradbury's response to Bill James' article "Underestimating the Fog". JC seems to think that James' article is misleading. I think he does make some good points about hypothesis testing, and as he says, it's something people wrestle with in academia all the time. But, early on in statistics or critical thinking classes, you're generally taught that you can't prove something doesn't exist. It seems fairly intuitive that, given past research on the subject, it may not be a very fruitful topic to look into at this point. My thoughts are the following (as I post at Sabernomics):
"I think the problem is that to find evidence for clutch hitting (even if there is such a thing), there would have to be some sort of inefficiency in managing strategies between the two clubs. If we assume there is a clutch hitting skill, then why wouldn’t there be a clutch pitching skill (I think you allude to something of that sort in your book, JC).
If that’s the case, then both managers should be optimizing their clutch from the defensive, as well as the offensive POV. Under this assumption, the results (or statistical data) would be a wash and there shouldn’t be significant evidence in the data for clutch hitting. If a manager doesn’t put in their so-called ‘clutchy’ guy when the other manager has in their ‘clutchy’ pitcher, then he isn’t optimizing his strategy. Unfortunately, for those trying to discern a ‘clutch’ skill, the only thing that would happen by putting in the ‘clutchy’ hitter is arriving back at the expectation that we originally had for the event.
But that doesn’t necessarily mean it’s not a repeatable skill. If, for some reason, there isn’t clutch pitching, then perhaps the idea of clutch hitting would be more manageable. Even if it were the case that it exists, I can only imagine it is quite tiny and probably not of interest in payroll as Jim states above. I think this ‘fog’ is just too thick for us to really find anything, and what could be found probably isn’t all that worth finding."
The main idea here is that, if managers are acting rationally to optimize their chances of winning in clutch situations, then we aren't likely to see a difference in the way the game plays out, on average. Why shouldn't there be clutch pitching? Someone like Mariano Rivera is referred to as a clutch pitcher. The problem is he's also just plain nasty. When a manager puts in a pinch hitter because of his clutchiness, the manager on the opposite side should be doing the same thing. This should just leave us with what we expected to happen in the first place, and doesn't leave much room for detection of anything. Any edge that a manger gets from putting in a hitter he thinks is 'clutchy' would just be offset by the other manager putting in his 'clutchy' pitcher--assuming the clutch factor is a constant increase in advantage across clutch players (or close enough to that). So while we don't detect anything, it could be an important part of the in game strategy.
With that said, I don't think there's much value in searching for the clutch factor. I think some players likely don't perform quite as well under pressure, but I'm fairly convinced that they get weeded out early on. There's always immense pressure to perform at the MLB level. If you can't handle pressure, you're not likely to be in the bigs. That's not to say that at a given point, a player in MLB doesn't fail due to pressure constraints. But on average, I don't think it's going to be anything influential over the course of the season. I'd be willing to bet that being 'unclutch' is a possibility from a psychological standpoint at lower levels of play. Unfortunately, I don't think we have the data to truly analyze that, and I don't think it's ever a reason to put a hitter in that, on average, isn't as skilled as the one already at bat.
Given a clutch situation, I'd love to have Derek Jeter at the plate. He's a damn good player and if I'm faced with that situation as a manager, I want my best player out there. If I had a choice between Jeter and Pujols, Bonds or A-Rod, well I wouldn't have to ponder too much. I'll take the guys that make hitting Major Leauge pitching look like child's play.
Addendum: I forgot to add some links about the discussion. Below are some other discussions, as well as a 'quick study' over at Sabernomics that gives a quick look at the 'effect' we might be trying to find.
Quick Study at Sabernomics
Discussion at Sabermetric Research Blog
The Blook Blog's Blurb
Note: I'm not sure why Tango thinks this isn't a 'practical problem'. Of course it is. It's trying to find the optimal way to manage, which makes it extremely practical (though probably not all that useful in the end). Just because teams don't currently rely on it, doesn't mean the motivations behind trying to find it aren't 'practical'. Maybe I'm being nitpicky with semantics, but this seems to be a point that Tango is really interested in making for some reason.