Wednesday, January 29, 2014

Computer Models-The Super Bowl and Econometrics

Andrew Isker emails:
I saw this article on Drudge and thought of a point you often make on EPJ about econometrics. 

Computer Simulation Predicts Seahawks Over Broncos, 24-21 - CBS Philly

No one in their right mind thinks this computer prediction is going to be accurate. Sure, maybe in a blind-squirrel-meets-nut sense it might happen, but no one is going to Vegas to make that bet as though this computer model is Biff Tannen’s sports book. 

A football game is only 60 minutes long, but it contains so many variables that affect the outcome of the game, both knowable (the relative skills of each player, the likelihood of injuries occurring, etc.) and unknowable (e.g. would Pete Carroll go for it on 4th and 1 on his own 45 down 10 with 8 minutes to play) that it is impossible to predict its outcome. Nobody would dispute this. 

 Now compare this with econometrics. Here you have computer models that try to explain what whole economies will do—whole economies, with hundreds of millions of people who each make thousands of economic decisions every day far more data than any computer model could collect, much less comprehend. 

In short if you wouldn’t find a computer prediction of a football game’s outcome credible, as you shouldn’t, why on earth would you trust anything an econometrist asserts?


  1. Mr. Wenzel,

    Don't you mean to say "The Big Game"?

  2. Ditto for global warming models.

  3. You're better off predicting doomsday all the time so that when something bad happens you can claim you were right. Hyperinflation due any day now.

  4. Great point Jerry... good thing there arent any government agencies or money printing agencies (i.e. The Federal Reserve) using Mr. Wenzel's analysis... come to think of it, if someone doesn't like Mr. Wenzel's prognostications they can just ignore him... unfortunately if a govt or money printing agency uses the Econometric's analysis we are all affected whether we like it or not.

  5. I used to work in casino in Las Vegas and supplemented my income by making 100 -200 dolars per week on college and NFL games in the sports books. They have syndicates that use the computer programs. As the article states NOBODY expected them to be right. The programs predictions were compared against the lines at the sportsbooks to guess whether the lines were going to up or go down. The sportsbooks lines were usually fairly good. You could make money bybetting both sides of the lines at different sportsbooks. If the game came out inbetween the highest and lowest bet spread you could double your money. If it fell outside you lost only the vig. No one could make money betting on the computers.