Friday, October 22, 2010

Finally, a Worthwhile Statistical Regression

A lot has been written on election theory and the influence of polls on voters. They’re self-reinforcing in two big ways: they discourage voters who support a trailing candidate (‘what’s the point in voting, the Reuters Poll says we’re going to lose by six points’), and sway undecided voters by identifying a frontrunner (appeal to popularity fallacy). I’d bet that the BCS polls have a similarly negative effect on the USA Today football rankings.

The first BCS standings came out Sunday. Boise State was supposed to reign over all, with TCU notching a lofty ranking as well. Yet the entire nation gasped in collective horror as Oklahoma was designated the greatest team in all the land, followed by the Ducks of Oregon. Now I know there are better teams than OU, and the fleshy polls do too. That’s why they’ve been ranked #5 or so all year. However, every ESPN article written over the past week regarding the Sooners identifies them as ‘#1 Oklahoma.’ This is wrong.

Assuming OU wins this weekend, they’re going to climb in the human polls. This will be defensible, as it will entail a victory over a legit top ten team in Missouri. Yet what if OU beats Colorado by nine points, and no other top ten teams lose? Or if OU has a bye? Would they still rise in the rankings – an effect of the #1 status bestowed via data processor?

It would be interesting to hold weekend performance constant and use BCS rankings as the dependent variable. Maybe somebody already has. With 12 years of usage, we don’t yet have a large pool to use. However, I have an inkling your beta level will indicate voters’ preference (or at least tendency) for making the team leading the BCS the recipient of more #1 votes.

Voters take their cues from other voters, who take cues from preseason rankings, which take their cues from the previous year’s success. Historical reputation creates an undue bias throughout. Polls are highly fallible due to voter reliance on cues and an inability to function in a vacuum. What’s worrisome is that the BCS rankings unduly enhance this foible.

Get On the Bus!

No comments:

Post a Comment