Thursday, March 20, 2003 :::
I'm trying to post this late enough that you can't steal my picks, but if you could, I'd like also to point out that I picked Texas Southern to win the anacrusis, so you don't want my picks. (I haven't posted my picks, but I may do that. I have submitted them to Kate and the Malcolms.)
I've paid less attention to college basketball this season than any time in recent memory; I paid occasional attention to the Ivy League when I thought Princeton could win it, but most of what I really know has come from half an hour of sports radio on Sunday night and some perusal of scores. I'll keep the discussion a bit away from issues specific to this year, then, though ignorance is certainly not a complete barrier to my willingness to discuss specific teams to some extent as well. (There is no discussion of the quantum topology problem introduced by the committee's BYU snafu.)
The bracket scoring, of course, is a bit different from what I'm used to; in past years when I've filled out a bracket, upsets were not given the Kate bonus, but the scoring also differed in that each game was worth twice as much as those in the round before it. The last three games of the tournament are worth half as much as the sixty preceding them, and if the team you pick to win the championship loses in the first round, you're out half your points in one blow (assuming, as has always been the case, that generally unspoken restrictions are applied). I don't like tails wagging dogs, so kudos to Kate on keeping the ratio sane.
This does, however, upset my normal strategy, which is to work backward; I start with who I think will win the tournament, then who I expect to lose to them in the finals, then I pick the other two regional winners, and so on. These picks are based partly on difficulty of region — if a team is in a difficult region, it's quite possible they won't be around for the championship — but it's mostly a top-down approach, and I've had decent success with it. This year the dog wags the tail, and the bulk of the points will be picked up in the mid-rounds; this is accentuated by the Kate bonus, because the seeds of teams tend to decline exponentially as the tournament progresses. This argues for the bottom-up approach that I expect most people use anyway, and it accommodates looking at match-ups, for which, as I said, I've come to this season particularly ill-equipped.
Now, 12-seeds are often the subject of derision, but this is an at-large seed; a team with a 12 seed is better than an NIT team, and is better than the sixth best team in a 5-berth conference, the fourth best in a 3-berth, etc. These are solid major-conference teams, not teams that should be in a national championship tournament, but teams that can beat low top-25 teams. 13 and 14 seed teams are NIT teams. 15 and 16 seed teams are cannon fodder. Based only on the seed, the 15 and 16 seeds aren't good enough, even with 16-1 odds, to justify picking a first round upset. Even if they were, that would remove the high seed from consideration for future rounds, which you might not want to do if the benefits are marginal. (On the other hand, if you're picking the higher seed to go out in the second round, there's no such cost, and you're back to considering only the odds of the first round matchup. An example: Syracuse against either Oklahoma State or Penn in the second round I expect to be sufficiently prone to an upset that I don't want to advance Syracuse farther than the second round. This leaves me to take the match against Manhattan on its own. On the other hand, where I might like SIU to beat Missouri, certainly with 6-11 odds, I think Missouri has a better chance of making it to the third round — they did just almost win the Big 12 tournament, and have a good tournament history — so I'll take Missouri in the first round because I have to. See also my national champion.)
It's well-established that people tend to estimate probabilities as closer to 50% than they really are; a terribly unlikely event will be given 1 in 4 odds, where the person guessing really would be a bit surprised if it happened one out of four times. I've maintained before that people are surprised by upsets in college football in particular more than I think they should be; this is kind of opposite to the other effect. My reconciliation of these is that I think people might be better at assigning probabilities to sports events than they are at other things, even if they are thereby underestimating how surprised they would be. Much of what I say is speculation, but this deserves its own special red flag; I have very little even crude anecdotal evidence for this, aside from the observation that top ten teams lose to unranked teams on a fairly regular basis and everyone seems to have forgotten the last time it happened. In any case, I've allowed myself, for filling out these brackets, to believe in the possibility of upsets. I probably should have gone out and actually researched how many of what seed have won in the first round, the second round, and so on; I largely just decided that a number of lesser teams were closely enough matched with their opponents that I'd hope for enough upset bonusses to compensate me for when I'm wrong. Thus I advance 3 12-seeds and 3 14-seeds to the second round, along with 2 13-seeds. In the next round, we give Roy Williams his exit sooner than we might anyway, and we pick Gonzaga, not because it's as good as Gonzaga, but because it's better than UCLA (and against a 1-seed, Kate's really pumped up those odds). We'll see how that works out for me.
Why, then, do I end up with Kentucky? For one, they assured me on sports radio that Kentucky is invincible. (See previous comments about forgetting about the last big upset.) It really seems to me that Illinois has a good chance of making the final four, but so does Duke, and Arizona, and even Kansas; I don't want to place that much faith in the team I pick to win that region. I see Kentucky or Pittsburgh coming out of their region, and probably Kentucky; if there were no final four, I might go with Pitt because of the 2-1 odds. As we get farther into the tournament, though, if we really believe the higher seed is better, you're asking for more and more stars to align when you pick, say, Louisville. That gets rough.
And that leads to a point I like to make, for example in a long post after the Fiesta Bowl: no team comes in with better than a 50% chance of winning a tournament this big; it is significantly more likely than not that the best team will not win. If we really wanted to know the best team, we wouldn't take several teams that have lost six of their last ten games and suggest that, if they win their next six, they must be better than the team that won sixteen in a row before losing the wrong game in overtime. But, of course, what we have is exciting, and isn't supposed to be sane; that's why they call it March Madness.
::: posted by dWj at 12:01 PM
