Brian's College Football Top 25

Since the college football season is upon us, it is time to unveil my phoney-baloney homemade computer poll of the Top 25 College Football Teams!

Top 25 for games played through August 31:

1 --- Ohio State
2 --- Oklahoma
3 --- Texas
4 --- Kansas State
5 --- Georgia
6 --- Iowa
7 --- Michigan
8 --- Miami
9 --- Colorado
10 -- Alabama
11 -- Marshall
12 -- USC
13 -- Texas Tech
14 -- NC State
15 -- Tennessee
16 -- Virginia Tech
17 -- Penn State
18 -- Florida State
19 -- Florida
20 -- Virginia
21 -- LSU
22 -- Wisconsin
23 -- Northern Illinois
24 -- Bowling Green
25 -- Texas A&M

It is the first week, and there are two more games going on today, so the model is still a bit funky and giving some odd results. The poll is odder still due to the particular biases incorporated into it- primarily that it ranks teams based on how many games they've won, when they won them, and what conference they play in. No complicated strength of schedule or margin of victory...

A more complete description of the model is as follows: A) The poll is based on rewarding teams that win, and emphasizing late-season over early-seaon victories. This means that on the flip side, losses early in the season aren't as debilitating as a loss late in the season. Each win gives a team points based on when the win was- the first game in the season is 0.5 pts, the last game in the regular season is 1.5 pts, such that Game 2 > Game 1, Game 3 > Game 2, etc. The scale is divided evenly among the number of regular season games. If a team plays more than the regular season amount (11 or 12), such as having played in a pre-season game or a championship game, a team would get progressively more points (essentially 1/12th or 1/11th extra per game, additive). A team's poll position derives mainly from the sum of all the "games won" points it has accumulated.

B) Each team has a conference, even the independents. The total number of wins of all the teams in a conference is divided to get Wins Per Team (WPT). The median WPT value is found among all 12 conferences, and a bonus or penalty determined based on how much above or below the median a conference's WPT is. This biases the poll against Notre Dame, which is good. This value is added to/subtracted from the total "win points."

C) A preseason ranking is used to give the top 50 teams bonus points that are added to their total. I used last year's model results plus some input from the 2003 Coaches Preseason Poll to determine the initial ranks. This bonus is added to the total.

D) Each team's record is compared to the average number of wins for all of Division 1-A. Teams doing better than average get bonus points, those below get a penalty.

So in summary, the model is: Total Points = (Win Points)+(Strength of Conference)+(Relative Wins)+(Preseason Bonus), and rank by current highest point total.

A big quirk of this model is that teams that "Play Early, and Often", are favored over teams that play later and have open dates. Since "Strength of Conference" (SOC) can fluctuate wildly based on the play of a team's conference mates, a team may win and *still* drop. A team that has an open date will probably drop, and teams who start their season later will be at a polling disadvantage during the season (catching up at the end). The model does not discriminate between wins over I-A foes or I-AA foes, or wins over quality teams vs. perennial losers- its "Just Win, Baby." I've found that adding in the SOC component has eliminated most of the anomalous results (with a team from a weak conference getting into the Top 10 because it ran the table against poor competition). Still, it's not impossible for say, Boise State, to pop into the Top 10 anyway, so long as they keep winning. By the end of the season, the driving variable will be wins, and anyone that wins out deserves recognition, poor conference or no. Since my model doesn't pretend to be predictive (except in a broad sense of 'those teams that win (and are thus highly ranked) are more likely to win over those that don't'), that's not really a problem.

If anyone wants to delve into the spreadsheet that I use to come up with this, I'll happily email it to you- comments & criticism are welcome.

Share this

Hmm, I'd like to see how

Hmm, I'd like to see how "my" team (North Texas or UNT) does as the season wears on. More often than not, the games appear to me to be punt fests. Unless we're playing a good team like OU or UT, and then it's a rout.

I do the rankings for every

I do the rankings for every DI-A team, should I post the extended rankings in the "more" section after the Top 25?

North Texas can't get a good offense going? Is UNT a private school, or just small?

No, I was just kidding

No, I was just kidding really about UNT. I mean, I assume you saw what happened Saturday? We played Oklahoma and lost 37-3. I don't know why we bother.

Meanwhile, one of our "rivals," New Mexico State, actually managed a TD against Texas in their 66-7 loss.

These teams aren't serious teams, why do they play the top-ranked teams in the country at all?

UNT is an enormous school, soon to be the second largest in Texas ahead of A&M. It's a public school and by far the biggest in the Dallas/Fort Worth area. Because of our pathetic football team, nobody's ever heard of the school outside of the golden triangle.

Well, for the teams that

Well, for the teams that aren't that good, they bother for the cash (the payout can be good), but if UNT is big, then they shouldn't be hurting that much for cash... the administration must not have any focus on football (or perhaps athletics in general?).

Oh, don't get me started on

Oh, don't get me started on that topic. ;)

There's a massive focus on football right now, actually.

http://www.nepentheisland.com/archives/000118.php

Hey. I'm wanting to make my

Hey. I'm wanting to make my own poll but have no idea where to start. If I could get my hands on an excel spreadsheet that I could tweak that would be great! One of my biggest problems is knowing how to handle all the data. If you're still willing to email your spreadsheet out I'd love it. Thanks!