Another year, another heartfelt attempt to start writing more regularly. But maybe this time it will stick. Certainly, sitting in the bleachers while my sweetie is doing laps, barracudas and various other maneuvers a lot of stuff wanders through my noggin. Hopefully the meditative space poolside will lead to more ideas, responses whatever. What I CAN say is that I am still ranking college football teams, and so how does 2024 look through 6 weeks?
The College Football Rankings for 2024
I have been doing college football rankings for a while. I’ve tinkered with the formula over the years, but I think I’ve landed on something pretty stable. Put simply, I wanted to make a college football ranking which would be systematic, objective and resume based. One of my real frustrations with how the sport does its business is how so often, teams exist as beauty pageant contestants. Notions of “best” vs “most deserving” permeate, though the two criteria are quite different.
Which team is “the best” is highly subjective of course. But it’s also amorphous, and so often the College Football Playof Committee has abandoned principles almost every year to satisfy what I can only surmise as “vibes”. Can a team have two losses and be better than an unbeaten team? Of course it can. Alabama was like that in 2022. But losses have to matter. Winning your league has to matter. Why do we play the season after all? In the other sports, the NCAA selection process can be weird, but it is largely consistent. College football is much more challenging, with many many fewer games - but it should be more transparent at least. (Play Fanfare) Here is where I come in.
The Inspiration
In previous years, I had used a computer algorithm based on the Bradley-Terry method to rank the teams. This was fine as far as it goes, but with so few games the error bars were high. The algorithm resulted in a lot of convariance based on certain nonconference results. It worked as a component, but it felt incomplete. So I looked to the NCAA Pairwise Comparison framework used for selecting college hockey teams. This framework compared all the teams with each other in terms of computer rating (RPI), head to head and record against common opponents. The school that “won” the most comparisons were ranked the highest.
The Application
To come up with the ranking, I compare teams across 4 dimensions:
Bradley-Terry ranking. I use a slightly modified version which assigns 0.8 wins to a team for winning the game and up to a 125% modifier based on margin of victory. (using a log-formula so there are diminishing marginal returns) The model also incorporates home field advantage. Having a better Bradley-Terry ranking is worth 1 comparison point.
Head to Head record. I give this pretty large emphasis. Winning a head to head match is worth 1.5 comparison point. In the unlikely event there is a rematch between two teams, a sweep could be worth 3 comparison points.
Common opponents. If a team has a better record against common opponents than another, it is worth a comparison point.
League champion. For now, this goes to the highest rated (see #1) team with the fewest number of conference losses. This is another comparison point.
After that, we sum up the comparison points and we get a head-to-head pairwise “win”. I make similar comparisons across all of the FBS teams, with a category called “Other” for all non-FBS opponents. Ties are counted as 0.5 wins. Ties are broken by using the Bradley-Terry ranking.
The First Rankings
I have moved the rankings to Google Docs, so you can see the whole file if you want. But here is the summary.
Notes
Personally, I have a hard time believing the SEC will remain as poor as they look so far. But Alabama's shocking lost at Vanderbilt has serious ripple effects, knocking down Georgia and knocking down Clemson. These things will reset I think as more games are played.
Army and Navy leading the AAC is fascinating right now. If they continue to dominate the league, they will meet in the AAC championship game the week before the annual Army-Navy game. This would be a pretty wild result.
Boise State is a bit high for my tastes. That said, with their single loss being a close game at Alabama, and right now Washington State's resume looking pretty darn good, I get it.
Comments