I watched the Yanks beat up on the Tigers last night. I’m floored by how much love is been giving to Derek Jeter (aka Mr. October, Captain Clutch) by the Fox crew. And again on ESPN.
Suppose we wanted to take a scientific approach in determining who is a “clutch” player. Obviously we’ll have to figure out what a “clutch” situation is, and then look at players’ performance in those situations.
A very important consideration though, and one that will be pointed out strongly and often when you take your QBA classes, is that sample size is extremely important in making statistical inferences.
I am pretty sure when you flip a coin that it comes up heads with probability very near 0.50. But if you flip a coin 4 times and if it comes up heads three times (75% of the time), is that strong evidence that the coin is loaded (that in fact heads come up more than 50% of the time)? The answer is no. A certain amount of random variation is to be expected.
In fact, in this situation, even with a coin that we were certain would come up heads with probability 50% of the time, we’d expect to see 3 or more heads 31.25% of the time if we flipped the coin 4 times. Now if we flipped a coin 400 times and sow more than 300 heads, that would be strong evidence that our coin is not “fair”.
Likewise, if you only saw last night’s game, Derek Jeter is the greatest baseball player in history. Of course, drawing inferences based on only 5 plate appearances is clearly dangerous if not entirely ridiculous. (Anyone remember Tuffy Rhodes?) Yet, if you listened to the broadcast, everyone knows Derek Jeter is “Captain Clutch”. How?
We could all disagree about exactly what is meant by “clutch”. Perhaps late in close games? Bottom of the ninth? Or simply in the playoffs? But even if we settle this disagreement, it is still very difficult to tell who is a clutch player, because by the nature of these situations, there are very few “clutch” situations during a season (or even a career).
Can we make inferences from these small sample sizes? Perhaps not with much confidence we are right. Just as flipping 4 coins will occasionally result in 3 or even 4 heads (is that “clutch” coin flipping or random variation), some players will may appear to be clutch players when in fact it is just random variation.
Is the case for Jeter? I don’t know. But neither do the announcers on Fox.
But for your amusement, I have taken Jeter’s stats and the much maligned Alex Rodriguez’s stats in all the playoff games they have played and listed them below. Because they have played different numbers of playoff games, I have adjusted their statistics – the numbers you see will are for each 100 playoff at bats they have. (Jeter’s sample size is much larger than A Rod’s). You can make your own conclusion. Could you do something fancier? Absolutely. You’ll see I purposely left off the names of the players.
Dr. Jahn Hakes and Dr. Skip Sauer (both economists at Clemson while I was in graduate school) have done some work on identifying clutch hitting in major league baseball. For an example of their work, click here. What do they find? They cannot find statistical evidence of persistently clutch hitters.
Why is that a couple of sports nuts economists, armed with PhDs from top schools, years of play by play data, and tons of computing power can’t find evidence of consistent clutch hitting, but the talking heads on Fox know it? Hmmmn.
If you like this type of analysis, then read a book called Moneyball by Michael Lewis. It’s about how the GM of the Oakland A’s (Billy Beane) listened to scouts (announcers?) less and started doing more statistical analysis in drafting baseball players. The A’s have been quite successful despite their relatively low payroll. It’s an awesome book for someone interested in baseball, statistics, economics, or even business in general.
–CT
I’ll post a comment later with the identities of the players. And for those of you looking for extra credit, bashing Jeter or A-Rod won’t do.
