Forget the polls and computers, let markets pick the top 2.
Eric Zitzewitz December 4th, 2006
Anyone who thinks that Florida is the second best team in the nation or would create the best matchup with Ohio State should check out the attached graphs:
First, note that on Saturday night, around the time of USC’s loss, there was if anything downward pressure on OSU to win the BCS, suggesting that USC was no stronger an opponent than some mixture of Michigan and Florida.
But then note how on Sunday, when the polls picked Florida (the computers had them essentially tied), Ohio State rallied significantly.
All this suggests that if Tradesports had a vote, it would vote:
1. Ohio State
2. Michigan
3. USC
4. Florida
This strikes me as a perfect application for decision markets. You run markets on the probability that a team wins conditional on being chosen, and give points to teams with the highest conditional prices. Heck, while we’re at it, we could do it for March Madness too.
The normal problem with interpreting these conditional probabilities causally is that the rest of the decision making process conditions on unobservable information, creating a selection bias. This doesn’t seem as much of a problem here — this is as close to common information as we’re likely to get in a decision making environment, and thus whatever uncertain component there is in the decision is likely to be uncorrelated with the teams’ strength.
Of course, given the NCAA’s longstanding problem with gambling, this idea is a complete non-starter. But it’s fun to think about.
Disclosure: The author is a lifelong Michigan fan who found the Florida coach’s incessant whining irritating but his comment that “it would be unfair to Ohio State” to make them play Michigan again on a neutral field deliciously (if unintentionally) ironic.
