BCS Championship Game Odds Model

This is my second week with my attempt to model the odds for teams to reach the BCS title game. Like last week, it is still a “beta” model, but much refinement has been made. I’ll offer my results first, then some insight into the methodology after.

Oregon	65.24%
Alabama	37.20%
TCU	27.25%
Nebraska	26.01%
Boise State	16.69%
Auburn	11.59%
Oklahoma	10.86%
"Other"	5.15%

Yes, this week it does indeed add up to 200%, what one would expect to get from 2 of these teams making the title game.

My problem modeling “other” here concerns who “other” is. Basically the way this works is thus – we know if Oregon and Auburn win out, they are going to the big dance. My model predicts Oregon’s chances of winning out, and Auburn’s chances, and comes up with a 5.51% chance of that game taking place.

What I do next is try to predict (it’s a model) who will most likely go to the BCS if the first matchup doesn’t occur. In this case, I chose Alabama. Although currently 6^th in the BCS, they have the NCAA’s toughest remaining schedule (LSU, Auburn, and SECC), and would likely take Auburn’s “place” if they defeat them. In other words, I’m predicting they would jump numbers 3-5 (TCU, Boise, Utah).

And so on. I also think a 1-loss Big 12 winner (Nebraska or Oklahoma) will jump the non-BCS 3. However, after that, I think TCU (currently 3^rd) and Boise (4^th) will get in if Nebraska and Oklahoma both fail.

As for “other”, it includes the unlikely Utes, and some Big Ten teams (Wisconsin?). For now I prefer leaving it as “other”, as all have such low likelihoods individually that it gets both messy and noisy.

As for actual match-up probabilities we have -

ALA-ORE	20.25%
NEB-ORE	12.82%
TCU-ORE	12.57%
BSU-ORE	7.86%
NEB-ALA	6.07%
AU-ORE	5.51%
TCU-ALA	5.15%
OK-ORE	4.78%
TCU-NEB	3.30%
BSU-ALA	2.69%
OK-ALA	2.50%
BSU-TCU	2.02%
NEB-AU	1.95%
TCU-AU	1.95%
TCU-OK	1.81%
BSU-NEB	1.54%
BSU-AU	1.24%
BSU-OK	0.87%
OK-AU	0.72%

So while I’m saying Oregon-Alabama is most likely at this point, it’s still only 1 in 5.

In making the model I am using the predictions of statfox, shown to have some of the most accurate predictions for NCAA football. After using statfox to calculate lines, I am using the overall record of Vegas lines as shown here.(Phil Steele)

statfox and Vegas seem to often be pretty close in their predictions of lines, even getting the line for Oregon over Washington exactly the same at 35.