Wednesday, October 03, 2007

BlogPoll - My Response

My resume ranking system takes some shots from our buddy Brian over at MGoBlog this week as being “just plain weird”.

If by weird it is meant “independent thinking” then we stand guilty as charged. Sure, I agree that perhaps Wisconsin at number 4, a team I don’t believe to be very good, is arguable, but the real issue goes back to what exactly does it mean to rank teams?

There is one school of thought in which rankers seem to feel they are engaging in a predictive process, as in “See, I told you Louisville was over-ranked”.

In addressing this school, let’s take a look at our collective genius efforts that resulted in the first preseason BlogPoll –

Michigan –WRONG!
West Virginia – WRONG!
Texas – WRONG!

And that is only spots 3, 4 and 5. Of the initial BlogPoll, the following initially ranked teams are unranked a mere 5 weeks (or 40%) into the season –

Penn State
Texas A&M

So, less than halfway into the season, 10 of the initial 25 teams we collectively thought in the top 25 are no longer there.

Not to mention the poll position falls for these still ranked teams –

Texas – 14 spots
Virginia Tech – 10 spots

There are others, but those stand out.

In other words, our initial BlogPoll sucked in having any predictive power at all. Likewise, I submit that the current poll will look equally bad by season’s end.

Another school of ranking thought, akin to the Predictors, is the Power Pollsters, or voters estimations of how good a team is. This seems to be the school of thought keeping USC near the top.

However if that was the overriding theme of the initial poll, well it sucked in that regard too. We quickly learned that Michigan, Louisville and Texas – all top ranked teams, were not very good.

(On another thought, I would like to see Brian run the “Swing” number between the initial and last BlogPoll. My bet is that “Swing” would win the manic depressive easily)

Finally, there are we “resume rankers” In my resume ranking system, the top teams are the undefeated teams plain and simple. I address this in my post about my rankings, and obviously I don’t think U Conn, Purdue or Kansas could prevail over many teams ranked higher. At this point, I argue it doesn’t really matter. Both my poll, and the BlogPoll at large, is far from a final product. If Kansas loses they will fall out of my top 25 quickly. And if they keep winning, they are going to go to a BCS game. At this point – and based on the prior history of both the “national” pollsters and the BlogPollers, whose to say they can’t? Don’t tell me your predictions to this point have been very accurate.

When I rank the teams, I approach it as a teacher grading a student’s test or paper. I may know very well that little Johnny Boilermaker is not a very bright student, but if he got all the answers correct, he is going to get an “A”. Likewise, I many know with a fair amount of certainty that Johnny Boilermaker is going to screw up a test or two along the way. When he does, he gets his “F” then.

As for my voting, when it came to the top, undefeated teams, I tried to rank them according to best win or wins. In my eyes, the 6 teams I rank ahead of USC all have better quality wins than USC does. USC’s marquee win – over Nebraska – isn’t that impressive in light of subsequent events (much like Kentucky over Louisville has become). But if little Janey Trojan keeps doing well on tests – as I suspect she will – she is going to move up.

Because it doesn’t really matter right now exactly where USC is ranked. If they are one of the two best teams by the time the season ends, it will show. If they lose, they probably won’t be. And not one of you voters can tell me with certainty USC isn’t going to lose.

You can quibble with me that Wisconsin’s wins are inferior to USC’s, and I’ll listen. You can argue that undefeated U Conn has an inferior resume to 1 loss Florida, and I can’t say you are wrong. But don’t tell me one team is “better” than another. You don’t really know that. You told me Michigan was the “best” Big 10 team this year, you told me West Virginia, or Rutgers, or Louisville, was BCS title bound, hell you even told me Ohio State was better than Florida last year. Keep telling me what is then proven empirically wrong, and I don’t believe you.

Plus, why should any of us care where any particular team is ranked right now, on any level? The AP doesn’t give out a trophy that says “Ranked 3rd in the 5th week of the season”. It only matters at the end.

Which is why resume ranking, to my thinking, is a superior product compared to the method used by the national pollsters. In resume ranking (in theory) any team can qualify for the mythical national championship game. We know that no team ranked lower than 20th preseason has played in the BCS title game since its inception. Right now South Florida, at 9th, has absolutely no shot if LSU and USC remain undefeated.

And that is a large part of what is wrong with the current system.


Year2 said...

In fairness, USF has no shot regardless if they and USC and LSU win out, because LSU and USC have tougher schedules and on the basis of resume ranking would end up ahead of the Bulls anyway.

KevinH said...

Here's a question about resume ranking. Say Florida beats LSU. Is it better to have a quality loss (LSU) or better to have a quality win (FL)? This is similar to the Canes over FSU in 2000 (in which they chose quality loss).

Clark said...

I agree with your basic premise that ranking teams should be like assigning grades. Look at what has actually happened rather than speculating about how good a team "really is". However, I see one flaw in your school test analogy. Jonny Boilermaker and Susy Trojan haven't been taking the same tests. If you and I are both getting perfect scores on math tests, but I'm taking Algebra I and you're taking calculus, who's better at math? And if you're getting Bs in calculus and I'm getting As in Algebra, am I better?

While it is certainly nebulous to try and use strength of schedule to assign value to wins, it is necessary. Otherwise, undefeated I-AA schools ought to be ranked. Or undefeated high school teams. Yes, it ruins the objectivity of your system, but there's little hope of removing subjectivity with 119 teams that play 13 games a piece.