Introducing the 2025 JAR College Football Ratings (v. 0.0.1)
Maybe I should have named it "Cornputer Rankings"?
Hello there. Been a while since anything posted to this site. If you’re curious why, I’ve explained it at the bottom of this page. Otherwise, enjoy this foray into rating every FBS teams in 2025.
I’ve become fascinated over the last few years with the dozens of rating systems used to evaluate college football teams. Some, like Colley, Sagarin, and Billingsley, were bona fide contributors to the BCS formula, existing long before the BCS was formed and still pumping out ratings to this day. Others ubiquitously appear in various media. ESPN publishes Bill Connelly’s SP+ rating each week (you know that one as the rating that always has Alabama rated higher than you think they should be), and Josh Pate consistently refers to “the model” he uses to make predictions. I’m personally a fan of Kelly Ford’s ratings, which are seemingly always floating around twitter.
I figured the best way to learn how these systems rate teams is to make my own. But if there are so many rating systems already available, is there really a point? Wouldn’t it be Just Another Rating?
The JAR Rating
It might be just another rating, but here it is: the very early, pre-beta release of the JAR rating. I’m going to throw the top 25 teams in the JAR ranking at you, and then we can talk about where these numbers came from.
I wanted a rating that would account for the Five Factors (not a novel idea—it’s what SP+ does), but how to do that in the preseason? I could use last year’s numbers, but that just gives me last year’s rating. I could try to adjust last year’s numbers for this year’s returning players—and that seems like something. In fact, for a preseason rating, returning production, talent level, and recent performance seem to be fair statistics to pull from. This preseason version of JAR is based on those three things: returning production (measured in usage rate and PPA), talent (measured in recruiting rating), and recent results (measured in Elo rating). Over the course of the season, the formula will shift to emphasize the Five Factors and actual production of the teams.
The number you see to the right of each team is how many points better than the average team this formula thinks you are (the average college football team has a rating of zero). I was surprised by some of the results: I would’ve expected LSU and Texas to be swapped, A&M to be a little lower, Florida to be a little higher, and… Auburn? What are they doing in the top 15? But overall, I don’t think this rating is too whack. I was happy to see Tennessee in the top 10, especially since I specifically didn’t rig the formula to be friendly to the Vols. Speaking of, how else can we use these numbers?
The Team Sheet
In what is probably a wild abuse of this rating’s capabilities, as well as an incredible show of hubris from an English major with a spreadsheet, I used JAR to calculate win probabilities for the season, and then ran 1000 simulations of each game. Based on those numbers, I had the formula declare each game a WIN, LOSS, or TOSS up. Those results for UT:
What you’re looking at here: top right corner, Tennessee’s JAR rating, rank, and projected record listed wins-losses (tossups). Down the schedule each week you see the opponent’s rank, name, and rating (remember the average team’s rating is zero). Then the last three columns are Tennessee’s win probability against that team, the % of wins from the simulations, and whether that game is a win, loss, or tossup. This is one spot I’m still doing some tinkering with where the thresholds are regarding what feels like a tossup vs. a win or loss. The UGA game, for example, has a win prob. under 50%, but the simulations were friendly, so it comes out a tossup. But you told me Tennessee’s record this year would be somewhere between 8-4 to 11-1, I’d probably say the ceiling seems high but the floor seems about right. BTW, I can run one of these for any team, so if you’d like to see a team’s results leave a comment below.
Going Forward
There are some flaws in the system. For one, I pull my data from collegefootballdata.com. Some of their spreadsheets don’t play nice. Team names aren’t always standard—sometimes it’s UConn and sometimes it’s Connecticut, for example (UMass and any directional school have a similar problem). Same goes for some players. DJ Lagway shows up in some spreadsheets as “DJ” and others as “D.J.”, which can throw some calculations off. In fact, I’m pretty sure Florida’s rating will go up if I go back and check those sheets. There’s also the fact that I don’t really know what I’m doing. I didn’t major in stats, I’m teaching myself most of this stuff with YouTube videos.
But overall, I’m pretty happy with how JAR turned out. I set out to create a system that would create a reasonable rating, and to my own astonishment I think I got there. I don’t think it’s better than the systems that inspired me. In fact, it’s probably demonstrably worse. I’m thinking of it more as a project or a process than a product. And I’m looking forward to seeing it evolve over the course of the season.
Extraneous
Why has this site been dormant the last few months? The bottom line is, every time I sat down to write about Tennessee basketball or baseball, I wished I was writing something about football instead. I absolutely love Vol basketball and baseball. But it turns out I don’t love writing about basketball and baseball in general.
When I re-launched Corn from a Jar a few years ago, the time seemed ripe for a heavy focus on Tennessee sports other than football. The athletic program in general is in what we’ll probably look back on as golden years, yet no other program gets the media coverage that football does. Why not carve out that niche? Be the guy that blogs all the other sports. It’s a solid plan that sounds good in theory, but I didn’t enjoy it.
So for now I’m sticking to football, despite the million other voices already in that space. After all, you can’t spell “saturated market” without UT.