In the hyper-competitive world of operations research blogging, needing to teach a class can put you hopelessly behind. The Blog-OR-sphere is abuzz with pointers to the CNN article on computer models for predicting success in the upcoming NCAA tournament featuring Joel Sokol (see the video here). See the blog entry at Punk Rock Operations Research as well as previous entries by me and Laura on LRMC (Logistic Regression/Markov Chain) and Laura’s article on the work of Sheldon Jacobson and Doug King. We previously saw Sheldon in articles on predicting the US Presidential election.
Getting back to the CNN article, it is a good illustration on how hard it is to write about models:
At their cores, the computer models all operate like question machines, said Jeff Sagarin, who has been doing computer ratings for USA Today since 1985.
Different people come up with different brackets because they’re asking different questions.
Sagarin’s equations ask three questions: “Who did you play, where did you play and what was the result of each specific game?” The computer keeps repeating those questions in an “infinite loop” until it comes up with a solid answer, he said.
Sagarin has arranged the formula as such partly because he thinks home-court advantage is a big deal in college basketball.
Other models ask different questions or give the questions different weights. Sokol, of Georgia Tech, for example, cares more about the win-margin than where the game was played.
Well… kinda. It is not that Joel has a philosophical belief in win-margin versus home court. It is simply that his models include win-margin and the resulting predictions are more accurate because they do so. Joel didn’t go in and say “Win margin is more important than home court”: it is the accuracy of the resulting predictions that gives that result. Some of his models don’t include win margin at all!
I also loved the quote:
Dan Shanoff, who blogs on sports at danshanoff.com, said gut feeling is more important than statistics, but taking a look at the numbers can never hurt.
Followup question: “So how do you know that gut feeling is more important than statistics, Dan?”. Reponse (presumably): “Well, it is really my gut feeling, you know, since I really haven’t looked at the numbers”. [Followup added: Dan isn’t sure he really said what he was quoted as saying.]
Be sure to check out Laura and me, and any other OR people twittering the tournament with tag #ncaa-or, starting noon Thursday.
Speaking of the tournament, I just saw the following interesting article complaining about how big-name schools tend to get more home games:
http://www.washingtonpost.com/wp-dyn/content/article/2009/03/21/AR2009032101994.html
This must be a difficult optimization problem, though, because some of the venues will be close to a lot of schools. Still, presumably you could do better than the current approach using combinatorial optimization. Thoughts?