Tuesday, March 18, 2003 :::
It looks like Kate won't let me fill out my bracket the way I'd like. If I think Penn has one upset in them, but not two, but that if Oklahoma State makes it past Penn they can cruise past Syracuse — let's be clear that my contempt for the Big East rivals that of the selection committee — then it makes clear sense to pick Penn to make it to the second round but Oklahoma State to make it to the third. It's a probability thing; I can't get a perfect score with an atavistic pick, but I hedge my bets and improve my most likely score.
Of course, even with that kind of freedom, how one wants to fill out one's bracket depends on one's goals. If I'm in a pool with 10 people and don't distinguish between second and tenth — I want to maximize my chance of first place, no matter what — I should take correlated risks that either pay off big or completely sink me. If I'm looking for a good "expected value" type of score, though, I play it more conservative. That's probably the way I'm going with this thing, in part because I don't know how many people I'd be trying to beat otherwise.
Incidentally, from that standpoint it's my experience that just picking the higher seed is a good strategy; that strategy typically beats most entrants but will never win, at least in my experience. Of course, with Kate's scoring system, that's all different. I'm not sure it's what I would have come up with, but it's different enough to be interesting to me; throw out a degree of the intuitive rule set you've developed for picking brackets over the years.
People have a certain amount of intuition for probabilities, but not too much, generally. On a not-unrelated note, the sum of shares of shared responsibilities shouldn't necessarily add up to one; if a number of people contributed to an event, the blame each person gets should be based on his marginal contribution to the probability that it was going to happen, and that can be less than 1 or greater than 1, easily by a factor of several times. That's not the way most people think, though; they see a pie of blame and want to partition it like something concrete. In cases of civil litigation, of course, there are practical reasons to pretend this is valid; if a victim is going to be compensated, the total damages awarded should come out near the loss, and courts can't simply pull money out of the air to correct a deficit, no matter how much anti-corporate juries might like to believe it. There's a similar problem on the surplus side, in that many actors may have contributed substantially to an event; the situation I can think of in which this practical constraint doesn't quite apply is covered by the notion of joint and several liability, where a tortfeaser with large pockets can be made to cover what a "judgment-proof" co-defendant is unable to pay. I don't entirely know what I'm talking about, but insofar as I do, the civil law seems to do okay on this front.
What was I talking about? Oh, probablility. Uh, go Packers.
::: posted by dWj at 1:25 PM