I just keep opening cans of worms, don't I? With last week's Top 25, I spent a good deal of time responding to questions, specifically regarding my ranking methodology. I gave an overview of the most common types of rankings, examined some of the strengths and weaknesses of each, and explained why I chose the method that I use. This, I thought, would settle the matter. I was wrong.
“Power polls,” or even power polls hybridized with resume polls, lose meaning as the season nears its conclusion. That is, the essential question of the power poll – “which team is better, and would win the most on a neutral field, right now?” – produces more and more intuitively objectionable anomalies as we reach December and January. As a result, it’s a methodology that comes close to but fails to telling us what we really want to know: what ranking does a team deserve?
My response partially rebutted his criticism, but it didn't solve the central quandry:
First of all, you state “what we really want to know: what ranking does a team deserve?” This is a loaded question. When you ask it, you are implicitly looking for a résumé ranking, as that is the question such a ranking is designed to answer; any power poll ranking will necessarily fall short of satisfying you. That’s not what a pure power poll is for. Of course Ohio State doesn’t deserve any ranking at all so far: yes, they’re 4-1, but those wins are against 1-AA Youngstown State, Ohio, Troy, and Minnesota. Get back to me after the Wisconsin game. Still, my eyes tell me that they’re still pretty good, so for now, they are given a middling ranking.
What we have here is a pretty fundamental conflict, centered around what question we should even be answering with a Top 25 ranking. Do we ask "Which team is best?" In such a case, we want a power poll. Or do we ask "Which team has accomplished the most?" Then we want a résumé ranking. In most professional sports, the schedules are fairly balanced, so the best teams ARE the ones that accomplish the most, but that is hardly the case with college football. The differences between such polls are real, and solid arguments can be made in favor of either approach.
So which to choose?
This question is not merely academic, as Calfan astutely points out:
Methodology for college football polling is particularly contentious because of the way we choose our champion. I won’t turn this into a playoffs vs. bowls debate, but the fact that the polls decide who plays for (or even who gets) the championship means votes have to do two things in tension: reward the teams with the most accomplished seasons and pick the teams playing the best at that time.
On one hand, schedules are not equal, and they are not entirely the fault of the schools. Those teams outside of the BCS conferences have basically no shot at accomplishing the most, simply because their schedule will not allow it. Utah in 2004 is a classic example. They blew away their entire schedule; just one team (Air Force) came within 2 touchdowns of them the entire year, and even the Zoomies lost by 14. To base things entirely on their résumé (especially given the BCS's margin-of-victory agnosticism) is unfair, and frankly, rather un-American. However, rewarding teams for finishing well is hardly fair either. Anyone think that USC deserved a shot at the big prize last year after dropping a home game to 41-point dog Stanford back in October? Yeah, neither do I.
Surprisingly, the AP, organizers of football's oldest and most respected poll, offer no guidance whatsoever on the issue. If memory serves me correctly, AP ballots merely ask voters to rank the top 25 teams in college football; specific criteria are left entirely up to the voter's discretion. The BlogPoll, for whom we compile our own Top 25, supplies more verbiage but not much more direction:
Preseason polls are supposed to be exclusively about how good a team is thought to be, and postseason polls are supposed to be exclusively about how much a team has accomplished on the field.
Now... it is impossible to separate the former from the latter in late-season polls because college football provides such a sparse data set, but at the very least BlogPoll voters know they shouldn't vote a 9-2 USC team #1 even if they think they're the best team unless that 9-2 includes three killer nonconference matchups.
I don't pretend that our ballot will have any impact on who college football's eventual national champion will be, but I still think a question such as this is worth discussing. After all, if we're going to bother to put together a Top 25, and you're going to bother to read it, you should be able to infer some sort of meaning from it.
So, to get back to the central question, what am I doing here? Honestly, it's much easier to say what I'm not doing. I'm not putting out a pure power poll, as I think ignoring a team's past record is both willfully ignorant and unduly rewarding of talented-but-inconsistent teams. However, neither am I putting out a pure résumé ranking, as I think ignoring what I've seen when watching all the football I see every Saturday is also willfully ignorant. College football has such a "sparse data set" that when trying to differentiate the top teams in college football, I need as much information as I can get, even if some of it is the sort of wishy-washy 'they just look better' evidence that I get via my cable package.
In essence, I am doing "two things in tension", and it is not simple. Yet, even as I admit it is a compromise solution, I attest that it is also the best one; neither extreme satisfies me. This leaves but one criticism left:
So if everything is a mess, why am I ragging on power polls in particular? Because at the end of the day, I don’t think they can stay internally consistent while rewarding the right things. What ragnaok said above – in describing his hybrid method – is just that on the continuum from power poll to resume, he probably shifts left to right over the course of the season. In any given week, though, a reader doesn’t know what the poll means – even with his disclosure of methodology.
He's got a point there. While either a Power Poll or a Résumé Ranking can be said to answer a specific question, my hybrid method is not so easily distilled. From week to week, it's a moving target, an evolving attempt at compromise. Still, that doesn't mean I'm making it all up from week to week; far from it. Every week I'm still ranking what I think are the top 25 teams. It's just that as more games get played, I gather more information -- more hard data -- and things I thought I knew are replaced by things I actually know. Is it a bit messy? Sure. However, I don't think I can come up with a list of 25 teams that I honestly think are the best by any simpler or more transparent method. If that doesn't quite satisfy you, well, I don't really blame you. Do you have a better idea?
Perhaps next week I'll talk more about why I think a pure résumé ranking is silly, or one of the other 3 or 4 topics I meant get to today, and didn't. Anyway, here's our Top 25 for the week:
This was a particularly divisive week to put together our Top 25. Frankly, Yellow Fever, CBKWit and myself didn't agree on much at all. Here's what we did agree on: Oklahoma is #1. BYU is #9. The other 7 teams listed here all fit together in some order between them. That's about it. USC puts its embarrassing loss to Oregon State behind it, both in their minds and in the minds of a lot of voters, in decisively crushing Oregon. Still, I think 'SC's got another loss in the somewhere down the road. Right now, I'm willing to call them the best one-loss team, for whatever that's worth.
Here's what else we agreed on. All 3 of us had all 9 of these teams somewhere on our ballots. Vandy moves up after a huge win over Auburn, although I remain skeptical, mostly because I think that little of Auburn's offense. At this point, I'm not sure how to differentiate teams like Oklahoma State and Utah; thoughts?
Also, Kansas falls despite winning. Not only does a 2-point win over Iowa State not inspire confidence, but remember their tight game @ South Florida earlier this year? Well, South Florida got punked by Pitt, so as the Bulls fall, they drag the Jayhawks down with them. Realistically, I think we should still switch South Florida and Kansas, at least until one of them takes another loss, but this is how the average works out. Come up with a convincing argument, and we'll change things 'round.
I think CBKWit said it best when he submitted his Top 25 to me: "Man, I hate everyone starting at 19. None of those teams should be ranked, but you have to rank somebody." Yep, here's a list of somebodies. Wild disagreement here. Not one of these teams showed up on all three of our ballots, as I think each of us struggled to find teams we felt deserved some measure of our confidence. TCU and Illinois are entirely my fault; I just didn't feel like rewarding undefeated teams with soft schedules, which is why I didn't vote for Northwestern or Ball State (who didn't make it).
And TCU? Sure, they got slammed by Oklahoma, but so has everyone else so far. And I challenge you to tell me how that loss is somehow worse than Ohio State's loss @ USC, yet everyone had no problem leaving the Buckeyes, who now have a comeback win over no-longer-ranked Wisconsin as their best win, in the middle of their ballots. I cry double standard.
What method should CGB use to come up with our weekly Top 25?
Power Poll (14 votes)
Résumé Ranking (11 votes)
A blend of the two (28 votes)
Some other method (explain in the comments section) (8 votes)
61 total votes