TOSCFR
(Posted to the site here.)
For the 2008 college football season TALB intends to embark on a quest to come up with a better ranking system. It is a “computer” ranking by definition. This is a work-in-progress, and is considered “open source”. Anyone can take this methodology, and fold/spindle/mutilate to make it better. And more specifically, a developer with the talent and the time can probably make this a reality more so than I could.
Why?
No one college football team plays enough common games with another outside of their conference that one could use those games to make an objective assessment of strength versus one another. The human polls attempt to use the human brain and eyeballs to determine which team is best. But humans are fraught with emotion and bias. Do you really think Steve Spurrier will ever vote Georgia #1? Then there are the computer polls. These remove bias, but since they are all closed we can’t know how they determine the rankings. I don’t necessarily disagree with any given human or computer ranking. This effort is intended to both remove subjectivity and to open the process to inspection (and real methodological improvement).
The Method
The rankings will apply only to teams in the Football Bowl Subdivision. Games against Football Championship Subdivision opponents will not be included or considered.
Pre-Season:
For each team, identify the number of returning starters.
The starting positions are the 22 offensive and defensive positions, the kicker, the punter, the kick returner, and the punt returner for a total of 26 possible players. Take the percentage of returning starters (e.g. 21 of 26 equals .808). Then multiply that number with the previous season’s number of wins (e.g. 9*.808 = 7.272). This number will be called the “Wins Projection”, though it should not be confused with any sort of prediction of how many wins a team might achieve in the following season. It is simply a starting point.
After four significant weeks of football (the rankings will publish the date ahead of time) the teams will have their win total added to their preseason WP. Thus if the example team above won all four games, their Current Wins would go to 11.272. The ranking going into week five would come from this.
Starting after week five, each team will be expected to win or lose depending on their ranking. And they will receive points based on the result.
1) A team is expected to:
a. Lose to teams ranked 3 positions above and higher by any margin
b. Win or Lose to teams ranked two positions above or below by 1-5 points
c. Defeat teams ranked 3-10 below by 5-15 points
d. Defeat teams ranked 11-20 below by 16-30 points
e. Defeat teams ranked 21 and lower below by more than 30 points
Meeting any of these expectations will not add or subtract from the current ranking.
2) A team will be awarded bonus points for:
a. Defeating any team ranked 3 to 5 above by any margin (+0.25)
b. Defeating any team ranked 6 to 10 above by any margin (+0.75)
c. Defeating any team ranked 11 or above by any margin (+1.00)
3) A team will have points deducted for:
a. Losing to teams ranked 1 or 2 below by any margin, OR failure to meet the scoring requirements from section 1 (-0.10)
b. Losing to teams ranked 3 to 5 below by any margin (-0.25)
c. Losing to teams ranked 6 to 10 below by any margin (-0.75)
d. Losing to teams ranked 11 or below by any margin (-1.00)
Let’s use our example team, Florida Tech and say that their preseason WP of 7.272 put them at #12. Then they opened with four wins, taking them to 11.272 and the #10 spot. If in week five they play #19 Western Georgia and beat them 21-17, they would actually lose 0.10 points since they failed to meet the scoring requirements from 1c, leaving them at 11.172 going into week six, which drops them one spot to #13. Then they find the #2 team on their schedule, and win the game. They would then have 1 point added under 2c, and take 12.172 into week seven along with a #9 ranking.
Comments, Improvements, Criticism, etc.
Initial Ranking
I know the initial ranking determination is a little sketchy. But I just can’t figure out a way to, well, figure out who is where based on zero games played. I settled on returning starters after first considering statistics. At that point I had to ask myself “which statistics”? Stats from the first 3 or 4 games? Too many teams schedule the local community college during the first few weeks of the season, so that’s out. Last year’s stats? That’s leaning on the achievements of last year’s team, which may or may not include any number of players this year. Which led me to the returning starters * last year’s wins point.
Please discuss.
Scoring Requirements
This might be controversial, as it obviously would encourage running up the score. But the only justification I can see for lowering the margins is political rather than methodological. If #3 Florida plays #78 Vanderbilt, the Gators darn well better beat the ‘Dores by four touchdowns. Please submit any methodological arguments for changing the margins.
Bonus Points
This is wide open. I’m willing to entertain any and all arguments for changing the position ranges and/or the point additions/subtractions. Including the scale.
FBS vs. FCS
This one is also open for discussion. My first thought was to give 0 bonus points for a win, and -2 or something for a loss. Then I thought, in order to encourage the schedule makers, that I’d make it -0.10 just for playing and -2 for a loss. But, that’s more political than methodological, so I tossed it. Now I’m back to no effect. But, and here it comes Wolverines fans, what about something like Michigan/App State? Michigan deserves to be penalized for losing to a team it should have handled. But how to penalize. Is -2 the correct penalty? I’m not set myself on how to proceed here, so I’ve left it at 0 either way.
Copyrights and License
Because I’d like to get credit should anyone actually run with this, I’ve protected the ranking methodology with a Creative Commons license. See below for details on how this can be used.
TALB Open Source College Football Ranking (TOSCFR) by
TALB is licensed under a
Creative Commons Attribution-Noncommercial-Share Alike 3.0 United States License.
