Today, I will be going over my latest model, a CFB Elo team rating model! This model/article was inspired by the amazing work from FiveThirtyEight, so if you haven’t had the chance, go take a look at their models!
For those who haven’t heard of Elo, it is a rating system created by a man named Arpad Elo to rank chess players. Today, it has grown to be one of the most used rating models in statistics and is utilized for many team sports, not just chess.
These models are useful because they grade teams based on their performance against other teams–they are able to make accurate predictions about each team this way, without the bias that a human might have.
How it works
- Each team is given a base rating of 1000 +team talent value gathered from 247sports.com recruitment data
- A team’s rating is adjusted each game based on the result of the game (win or loss)
- A win against a very good team gives a team a larger Elo rating increase than a win against a very poor team
- A loss against a very poor team gives a team a larger Elo rating decrease than a loss against a very good team
- The margin of victory/defeat also affects the magnitude of the Elo rating change
- A team’s rating is adjusted each game based on performance (EPA for and against, yards gained/allowed, etc.)
- Using each team’s schedule, we can simulate the season 1,000 times and predict the chances that…
- They will have a CFP Bid: I defined this as losing 2 or fewer games
- They will be Bowl eligible: Win 7 games
With all that said, let’s jump into our preseason rankings, (which are based on 2021 stats). Over the course of the season, the model will run and update its ratings to follow the season live.
Find the full graphic of ratings for all 214 teams here (sorry it has to be an external link, blame WordPress for not allowing HTML file inserts!). You will notice some teams’ records are not their full season records–this is because our data only covers FBS teams, so non-FBS teams might be thrown in the system for a game or two if they play an out-of-conference game. Also, note that the ‘CFP Bid’ column is a predicting for the 2022 season using the Elo ratings from 2021.
Things to note:
- Two non-Power-5 teams finished in the top 5! This rating system doesn’t discriminate!
- Alabama was the only CFP team not in the top 4 ranked teams to finish the season–this can be accredited to their big loss to Texas A&M, which cost them -83 rating points.
- Wisconsin (9-4) finished the season ranked 9, after beating some strong opponents in the back half of the year–can they ride this momentum into 2022?
- Some surprising teams to make the top 25: Central Michigan, Western Kentucky, Fresno State, Army
- Some teams I was surprised to miss the top 25: BYU, Oregon, UNC, Auburn
- Shout out to last place, UMass! Only up from there!
Be sure to utilize the search bar if you are looking for a specific team!
Additionally, here is the static table of the top 25 teams:
We can also look at how teams’ ratings progress over the course of the season. See some examples below.
- Bama, Oklahoma, Ole Miss stuck near the NCAA Best teams all season long
- Auburn started strong and finished the year below average!
- Michigan and Cincy started and ended with bang!
- Wisconsin’s mid-season trouble turned around with quite the late rally as they finished just above average
- Indiana’s season went down the drain after a somewhat hopeful start
- Texas and Miami had mediocre seasons the whole way through
- Kentucky rallied early and finished strong to secure the 18th rank in the league
- The Ragin’ Cajuns had a season-long tear!
- Florida had an average season year-round
- Duke, USC, and Washington all disappointed big-time!
That’s all for my CFB Rating system preview! Stay tuned throughout this season for weekly ranking updates!