Friday, October 29, 2010

BCS Championship Game Odds Model

Based on some of the things I’ve been working with recently, I’ve created a model that attempts to estimate the odds of various teams reaching the BCS title game from this point in the season.

This is very much a “beta” version, if you will. The intent is to refigure these odds after each weekend’s games, updating them to reflect what has occurred. As such it should narrow as we approach the end of November.

I’m on my way to Jacksonville today, so I don’t have the time or energy to reveal all the inner workings right now. The model utilizes Jeff Sagarin’s Predictor, as well as the odds of winning for favorites based on Vegas lines. Also, I am an admirer of the work of Nate Silver at fivethirtyeight.com (yeah, politics), and his work inspired some of my modeling.

At this point I am only modeling teams that have what I will call an “unimpeded” route to the BCS title game. This would include Alabama, but not LSU, as LSU no longer controls their own destiny.

I have also made certain assumptions based on current BCS standings. Right now, it’s pretty obvious that if Auburn and Oregon win out, that will be the BCS title game. However, should Auburn lose, my model then goes to the next contender (in this case that’s Michigan State). I’m basing my selection of “next contender” on my knowledge of the BCS, including the computer polls. There is some subjectivity here, but I’m reasonably sure that a Spartan team that wins out goes to the title game over a Boise team that does the same.

Finally, I’m considering any loss at this point as disqualifying. This isn’t necessarily true, but I think it is truer than not at this point in the season.

Here’s what the model is showing -

Percentage Odds Team Plays in the BCS Title Game




Auburn
8.01%
Oklahoma
8.60%
Nebraska
10.66%
Missouri
18.03%
Michigan State
22.07%
Oregon
22.81%
Alabama
24.16%
Boise State
26.67%
"Other"
27.91%

First of all, this doesn’t add up to 100%, because there will be 2 teams there.

And yes, there’s Boise leading the pack, even though I have previously written they aren’t going to make it. So sue me.

Without peeling back all the considerable math that went into this, here’s what we can read about this model –

1. Auburn has a tough schedule ahead with Alabama and a prospective SEC title game. Sagarin has them as a considerable underdog to ‘Bama.

2. Alabama, by contrast, should be favored in all remaining games. If they win out, they should go.

3. Oklahoma and Nebraska aren’t out of this, and if they win out, they should go.

4. Boise has the best chance of running the table. If they do, and factoring for most of these other teams to lose at least one, this is their chance of playing for the BCS title.

5. At this point, the model has “other” as a 28% chance. Unsatisfying, yes, but it will change quite a bit after this weekend (TCU is probably quite a bit of that 28%).

6. One of the factors this model accounts for is games remaining. This is extremely important. Michigan State, for instance, has only 4 games remaining, so their odds are higher than one might expect, and higher than a team with more games or the prospect of a championship game.

The model also shows the odds of potential BCS title game match ups –

BSU-ALA
5.50%
BSU-ORE
5.25%
ALA-ORE
5.17%
BSU-MSU
4.43%
MSU-ORE
4.38%
ALA-MSU
4.36%
BSU-MIZZ
4.34%
ALA-MIZZ
4.27%
MIZZ-ORE
4.23%
MIZZ-MSU
3.60%
BSU-NEB
2.81%
NEB-ALA
2.67%
NEB-ORE
2.31%
OK-ALA
2.19%
BSU-OK
2.19%
BSU-AU
2.16%
NEB-MSU
2.00%
AU-ORE
1.90%
OK-ORE
1.88%
MSU-AU
1.66%
OK-MSU
1.64%
MIZZ-AU
1.60%
NEB-AU
0.87%
OK-AU
0.70%


Yes, that’s Boise State against Alabama at the top, but please realize that there is a 94.5% chance of this not happening per our model.

The real purpose of doing this model is to promote thought and discussion, not predict with pinpoint accuracy some odds. Alabama, for instance, is modeled to have a 1 in 4 chance of playing for the title. In other words, don’t overlook them as having a shot at a repeat.

Finally, this is going to get a whole lot clearer after this weekend. Many of these teams have one of their toughest remaining games this weekend, including –

Oregon at USC
Missouri at Nebraska
Michigan State at Iowa
Alabama at LSU

Any of these that clear this weekend’s “hurdle game” (particularly Michigan State) will have much higher odds come Monday of making it to the BCS Championship.

2 comments:

Amos said...

I'm curious about the math behind this. Two comments, shouldn't the percentages add up to 200% since like you said there will be two teams there?

If using Sagarin's predictor for determining favorites, shouldn't you also use the statistics of how accurate he is in that, as opposed to transferring that over to Vegas odds? I remember in the past you would "bet" some games using the predictor vs the odds, though this would be different as it's the predictor's ability straight up.

Mergz said...

Amos,

Still work to do, and I'll try to find some time to post on methodology next week.

I use the Vegas odds for the current week, and Sagarin for the future (he's usually pretty close to Vegas for the current week). I'd love to use something else for future games, but what?