Computer v. the Experts v. the Sports Proletariat
Popular consensus has been much revered these past ten years, popularized by Smart Mobs and The Wisdom of the Crowds. It has even spilled over into our political discussions, bipartisanship being the resulting much sought holy grail. The idea is that you'll get a pretty accurate prediction or estimate on something if you just let a whole lot of people vote on it. The same principle has been popularly applied to market theory (hey, they can't all be wrong, can they?) I've always felt the concept was a lot of mush, nothing more than popular anti-elitist esteem building to push back against disciplined statistical analysis.
College football, of course, has been way ahead of the curve, basing the process of selecting its national champions on polls. But there are polls and there are polls. There's the USA Today Coaches Poll, the AP Sportswriters poll and the various computer polls. The results of the first two are suspect, as I pointed out in a previous post, subject to regional bias. Are the computer polls any better? Inquiring minds want to know.
I thought I'd do a quick check, picking the well-regarding Sagarin poll and comparing its bowl predictions to the AP, USA Today and point spreads (which are determined largely by the betting public). Given that the computer poll can sometimes produce some eye-popping, seemingly preposterous conclusions (e.g. Stanford is rated #1 in Sagarin's "predictor" analysis), the results are enlightening.
On my sampling of a dozen bowls involving major conferences, the AP and Coaches poll both collected 7-5 records, wrongly picking winners in the same games: South Carolina, Michigan State, Missouri, Nebraska and West Virginia. Sagarin's computer poll and the point spread odds both went 9-3, correctly picking Florida State over South Carolina and Alabama over Michigan State (the latter is the poster child of how ridiculously biased the polls can be).
Based on this sampling, the computer bettered the "experts." But so did the point spreads, determined as they are by the unwashed masses. There's going to be separation in the next two major bowls, however. The AP poll, USA Today poll and the mob all pick Auburn over Oregon and LSU over Texas A&M. The computer has LSU and Texas A&M in a dead heat. But only the computer picks Oregon. We shall see...