Post date: Jul 20, 2011 8:12:5 PM
Originally published October 10, 2010.
So as I was glancing over the most recent college football rankings, I found myself asking the same question that always crosses my mind when reading those all-too-important numbers, "What the hell are these people thinking?". In an effort to determine just how accurate the supposed "experts" are at determining the best teams in the nation, I decided to take a scientific approach to address my question. My methods are, admittedly, EXTREMELY simple, but I think that they reveal some interesting trends nonetheless. Basically, all I did was determine (1) the overall winning percentage of each team's opponents and (2) the overall winning percentage of each team's opponents within each respective opponent's conference. (1) was a fairly obvious metric, while (2) was a way I could correct for possible artificial inflation in a team's opponents' overall winning percentage due to the tendency of teams in college football to schedule out of conference games that actually more resemble glorified scrimmages. Once these winning percentages were determined, I ranked each team in each category and then averaged those rankings together. I then ranked the teams based on their averaged rankings. Next, I calculated the difference between a team's AP ranking and my averaged ranking in an effort to determine the extent that the AP poll was under-ranking or over-ranking different teams. I felt the numbers for some teams were slightly boosted because their opponents had unusually high records as a result of playing in (what I feel are) weak conferences. I corrected for this artificial inflation by removing those opponents I deemed unsatisfactory and recalculating the values and rankings for the applicable teams. This decision is, obviously, one of personal preference and can be debated.
What I found was really interesting. The new top 10, in order, was:
Oklahoma
LSU
Boise State
Nebraska (tie)
TCU (tie)
Auburn
Oregon (tie)
Ohio State (tie)
South Carolina
Alabama
According to my ranking method, Oklahoma (AP:6, Mine:1) and LSU (9:2) are being substantially screwed by the voters, while Oregon (2:7) and Ohio State (1:7) are receiving a significant boost. You could justify LSU's low AP ranking by arguing that they have been getting fairly lucky in recent games, but the discrepancy with Oklahoma doesn't seem to have an explanation. Worth noting is that Auburn only moved up one spot in my system, so I feel like I did a pretty good job of keeping my personal bias out of it. That being said, I don't think Boise State and TCU actually merit their rankings in my system, so that would mean moving Auburn up to number 4 in my eyes. Obviously, I think Ohio State and Oregon's AP rankings of 1 and 2, respectively, are completely absurd. They actually finished dead last in my raw data rankings, but I had to, on principle, put them ahead of one-loss South Carolina and Alabama.
To conclude, it seems like the voters are ranking teams based on hype and convention, rather than performance. This leads to a blatant misrepresentation of the best teams in the nation to the general public. Opponent record data, like I analyzed, along with other similarly relevant statistics, should be readily available alongside published rankings so that the voters can be held accountable for their decisions by the millions of college football fans. Of course, none of this would matter with a playoff system.
That's the end of my rant. Feel free to check out my data and fact check me, in addition to offering any comments and criticisms you may have!