Monday, January 10, 2011

Championship Game

Tonight's the big night, the championship (or at least as close as there is to one) for college football. The Oregon Ducks are statistically maybe the best team in the nation and have been at the top all season. The Auburn Tigers have their doubters, but they also have the Heisman Trophy winner and they're favored by most at places like ESPN. So who's going to win? There's more at stake in that question than you might think.
On one hand Oregon is heavily favored by both of my rankings systems, having a 9 point advantage and 2/3 chance of winning. This is the same system that currently places in the 98.9 percentile in ESPN's College Bowl Mania. A closer look though reveals two teams going in opposite directions. Auburn started off squeaking through some close games against poor teams, while Oregon thoroughly trounced all comers, including a 21 point win over Stanford. Since then though, the Tigers have trended sharply up, with the Ducks have gotten worse. In fact, continuing the trend lines out to the championship game the Tigers are suddenly favored by 10 points! Here's a look at the game by game performance:

Auburn Tigers Game Score By Game

Oregon Ducks Game Score By Game
Which of the two systems should we go with, and how do such different outcomes fit together? One key trait of both ranking systems is that they count all games equally, regardless of when they were played. The assumption of consistent skill ignores the possibility that a team like Auburn could improve over the course of the season. On the other hand, relying only on the most recent games leaves the possibility that one fluke game would have too strong of an impact. Especially with 12 game schedules I think it's important to take all games equally into account to cancel out some natural variation. Uncle Leo might even note that this method has led to a 58% success rate against the spread in the NFL this year (but that's another article). So tonight I'm going to stick with my prediction that Oregon wins by 9 (41 to 32), but if it goes the other way it might be time to rework the rankings systems. Or it might just be random variation.

1 comment: