Monday, October 13, 2008

Long rambling post about polls, championships and the irony of Saurian selectors

When I approached Mergz about starting this blog I knew it was going to be fun and stimulating based on our previous lives as posters on UF message boards. While I might have a bone to pick every now and then with Mergz theories I generally agree with him. He attempts to make some sense out of thousands of variables in modern big time college football using rational arguments and math. That doesn't mean he's infallible. But if you're going to attack a position you've got to come with something more than "Team A is clearly better than Team B".

When we started this blog we knew we both agreed that the biggest fundamental flaw in college football is the National Championship along with the bowls & polls method used to determine it. Instead of having games that remove all doubt champions are crowned based on opinions. Thus Mergz and I don't have any respect or fondness for the BCS Championship despite the fact that our beloved Gators won it two seasons ago.

The irony is that we have now become somewhat complicit in the current mess that is college football. Last year we joined the blog poll and Mergz joked that we had now become "selectors." There's literally dozens of self-appointed selectors and the A.P. and U.S.A. Today voters are perhaps the best known. To add to the irony we've been given a degree of legitimacy as selectors now that the blog poll has been incorporated into CBSsports.com's web site.

So now our blogpoll selections are serious business. Some readers come in and don't understand the background or the shifting methodologies as the system progresses and thus comment ignorantly while others are trying to follow along and really provide concrete feedback.

Now it should be known that I've only met Mergz in person once and we don't talk about the blog much offline. I simply follow what he's doing and for the most part agree with it. For example, we have to submit a preseason ballot for the blog poll. How IN THE HELL are you supposed to rank teams that haven't played yet? Well Mergz' answer is to do it based on perceived quality of rosters as based on recruiting rankings. This year he added a tweak to it: bonus points for teams that have a recent history of turning star recruits into winning teams.

Then the games begin and limited information starts to come in. Remember when ECU looked like they were going to be the BCS busters this year? As we get more information about the teams based on wins and losses the picture becomes clearer. But we're still only halfway into the season and less than halfway into most teams' conference schedules. So there's a lot left to learn.

One of the biggest problems in ranking teams is contradictory information. How do you handle two teams with identical records in which one team beat the other head to head? Logic tells you that the head to head winner should be ranked ahead. But what if that head to head winner lost to a team with a terrible record? It's not just us that have this problem. Every selector has to make arbitrary decisions about these sort of things and as Mergz says most of them probably don't take it as seriously as he does. I mean do you think the AD's that fill the coaches ballots really scrutinize every result of every team they are ranking and trying to make sense of them?

The other big problem is that there isn't agreement on what the rankings are supposed to do. Some people believe they should be predictive: "Team A is ranked higher than Team B therefore they should win." There's nothing wrong with doing it that way except that you are often going to be wrong. Mergz has a "power poll" which he rolls out that attempts to do this. He uses team statistics to determine which teams are actually better. The trick is to have enough data for it to be meaningful (enough game results) and also to pick the right stats that have the most to do with winning.

But for our blog poll ballot Mergz has chosen a resume system. It's about who deserves to be ranked where based on who they have beat and who they have lost to. This is an important distinction because if instead of the blog poll we were voting in the Harris Interactive poll we'd be part of the process by which TWO teams are SELECTED to compete in what masquerades as a championship in a sport with 119 participating teams. If you have two job openings and 119 applicants are you going to choose the candidates that have a track record of accomplishment or the ones that have a lot of "buzz" and "potential"? We've determined that we would want as much empirical data as possible about how much the candidates have really accomplished.

Here's a question for you. Let's say there's an undefeated team ranked number 1 in the country that is led by an all-world QB and an all-world WR (let's call them Tebow and Harvin). In the second half of the team's last game both star players suffer season ending injuries. The opponent comes back and only loses by a couple of points. Now the value of the number 1 team is seriously in question because the back-ups simply are not as good. Should this undefeated team get the nod for the BCS championship game? If you were trying to be predictive, you'd say that this team's chances of beating some of the other highly ranked teams are now severely limited. If you downgraded the number 1 team because you felt that in their current state that they could not possibly beat some teams below them you'd be making a purely speculative decision. A resume ranker would say, well those are the breaks. The team beat everyone they played and thus deserve to play in the big game even though you might know intellectually that they have a very slim chance of winning.

What's the right answer? If you think about it, in every other sport there's a resume system. If Eli Manning had gotten hurt during the NFC championship game and the Giants had still won, they would still have represented the NFC in the Super Bowl.

Resume ranking is merely an attempt to substitute for a playoff and trying to determine who deserves it more. It's still flawed and stupid and often contradictory but it's the best we got.

You simply can not have a 2-team playoff in a sport with 119 teams, that play only 12-13 games each, and expect it to have any legitimacy. The answer is a playoff in which all 119 teams have a chance to win their way in. Ten 12-team conferences with divisional play and conference championship games is necessary. We can argue about seeding the 10 teams and who gets a bye but a large part of the need to have polls would be gone.

6 comments:

Amos said...

Great post, I agree with everything except the reason "on the idea that past performance is the best indicator of future performance".
I think that shouldn't even play into why the resume ranking is used. You get ranked high because you deserve it due to what you've done. Saying anything about future performance is now going into all those things you don't want to go into (like injuries and who's actually better).

Gator Boys said...

You've only met Mergz once??? I've ALMOST met him that many times.

Also I've always wondered how to pronounce the name of the site. Is it SORE-IAN suh-gas-sity or SAGA CITY?

Also, Mergz do you get a Keg in your van?

Gator Boys said...

I like how you do resume voting, I think its the best way. I have a question though. Hypothetically if LSU beats Alabama next week (if they played) don't you have to do a bit of subjective decision making? You could either say, "Alabama is a really good team, I guess I underestimated LSU" or "Alabama lost to a pretty bad LSU team, I guess both Alabama and LSU are pretty bad."

I assume that as the season goes on you re-adjust how much quality wins and losses get right? If LSU wins out and goes to the SEC championship game then Florida will go up a little bit because their win over LSU was better. Right?

Henry Louis Gomez said...

Amos,

As I wrote that line I was regretting it. But I was thinking about the ultimate reason we rank teams, to select a champion. I took the job interview scenario to it's logical conclusion. The reason you look at a resume is because you want the best possible hire. I will update the post because you are right.

kevinh said...

You're what's need in the polls: An independent, non-biased "show me what you've done" ranking system. It's not that I agree with you on all fronts, but that doesn't matter. It's the fact that you don't try to agree with everybody else. There is no "correct" system with polls, but, in the main stream media, there is only one correct system: "I think team A would beat team B."

DolFan 316 said...

I personally think the "ranking of the rankings" poll was extremely accurate and would like to see that done on a regular basis. And when you think about it, ranking teams according to their own rankings in scoring offense and defense is resume based as well. What better statement of who a team is than going by their points scored and allowed?

But I do like that you guys are at least thinking outside the box. My capacity to do this as far as ranking college teams is limited because I have a tendency to consider games in a vacuum.