(non-election) polls and margins of victory

Earlier this week was a detailed post. Today a specific example of one problem with lazy poll voters in college football: the margin of victory. Quite simply, margin of victory in itself is a relatively useless bit of information. Every game is unique.

Using margin of victory creates hilariously ridiculous constructs.

For example, margin of victory is one of the few things that people who rank a team like Boise State over a team like Missouri or Auburn can point to as evidence for why Boise State is a 'better' team. Well, here's a little fun comparison. Colorado is either the second to worst or worst team in the Big 12 (that will be decided when CU plays KU). This Colorado team beat Hawaii 31-13, an 18 point spread.

That's the same Hawaii that's tied for first in the WAC, Boise State's conference.

Now, Hawaii and Boise State have both played Louisiana Tech, another WAC team. Hawaii won by 20 points, and Boise State by 29. Using margin of victory, that makes Boise State look a little more impressive than Hawaii. How much more impressive does that make them look than Colorado? Then, if you learn that Missouri beat Colorado by 26 points, how would you compare Boise State and Missouri? This sort of exercise wouldn't suggest that Missouri and Boise State are roughly equal teams. It would suggest that Colorado and Boise State are roughly equal teams, with Missouri being notably better.

Now, I don't think Missouri is significantly superior to Boise State; I think they're pretty evenly matched teams, actually. But if you're going to apply concepts other than wins/losses and strength of schedule, then you have to be prepared to accept that sometimes, the additional factors can make things even more bizarre. If you subtract Colorado's 18 point spread from Boise State's points, that makes their margins of victory look a lot less impressive.

Expecting schools like Michigan State, Missouri, or Auburn to post point spreads remotely close to Boise State (or Oregon, for that matter, whose only game against a team with a winning record so far is Stanford) is comical.

Fortunately, the season is long enough that if the Spartans, Tigers, and Tigers continue winning, they will jump Oregon and Boise State. But particularly with Boise State, the question is, why wouldn't they already have jumped them? They've got the same record. They've beaten better opponents. And margin of victory looks pretty similar when the games in question are actually analyzed. All three of the Big 12/SEC/Big Ten undefeateds have shown they can win by over 20 points. They just can't do that regularly against top opponents.

After all, just look at the score from the Boise State - Virginia Tech game. Boise State won by a whopping three points, less than Michigan State's ten point victory over Wisconsin, less than Missouri's nine point victory over Oklahoma, and less than Auburn's seven point victory over LSU. Both the human pollsters and the computers agree that Virginia Tech is not as good as Wisconsin, Oklahoma, and LSU.

I think it's highly unlikely that all three of these teams remain undefeated. But at least one of them might, and it will be really interesting at that juncture to hear why people place Oregon, Boise State, and/or some one-loss team (say, Alabama) ahead of them.