The College Football Playoff selection committee released its first rankings of the season on Tuesday night. Miami is No. 1, which is new for the program. Most of the list looks familiar. Two placements do not: Notre Dame at 11 and Ohio State at 12, both one bad week from falling out of a 12-team field. (CFP, NCAA.com)
Our model has those two teams first and second.
That gap is what this piece is about. The committee ranks résumés: who you beat, who you lost to, and how it looked. Our model rates teams play by play, adjusted for opponent, and blends in a preseason estimate that fades as the season goes on. Neither approach is the right one. They answer different questions, and the places where they disagree are the most interesting games left on the schedule.
Every number below comes from our public college football ratings and record, as of Friday morning, November 6, after Week 10.
Our top 12
| Rank | Team | Record | Offense rank | Defense rank | Schedule rank |
|---|---|---|---|---|---|
| 1 | Notre Dame | 7-1 | 2 | 2 | 65 |
| 2 | Ohio State | 6-2 | 3 | 3 | 13 |
| 3 | LSU | 6-2 | 19 | 1 | 25 |
| 4 | Indiana | 8-1 | 7 | 4 | 36 |
| 5 | Miami | 8-0 | 1 | 32 | 45 |
| 6 | Georgia | 7-1 | 4 | 8 | 51 |
| 7 | Oregon | 5-3 | 18 | 5 | 16 |
| 8 | Alabama | 6-2 | 11 | 13 | 6 |
| 9 | Utah | 8-0 | 8 | 26 | 56 |
| 10 | BYU | 7-1 | 20 | 16 | 35 |
| 11 | Wake Forest | 7-1 | 9 | 28 | 72 |
| 12 | Texas Tech | 7-1 | 29 | 6 | 59 |
Offense and defense ranks are opponent-adjusted efficiency among every team we rate. Schedule rank is the average rating of every opponent on the full schedule, where 1 is the hardest. The full table, sortable, is on the college football teams page.
Eight of our twelve also appear in the committee's twelve. The four that don't are LSU, Oregon, Wake Forest and Texas Tech. The committee's four that aren't in ours are Kentucky, Texas, UCLA and Ole Miss.
Where we disagree
Notre Dame and Ohio State: the model's top two, the committee's last two in
These are the only two teams in the country that rank top three on both offense and defense in our numbers. Notre Dame is second and second. Ohio State is third and third. The model reads that as the profile of the best teams in the sport, whatever the loss column says.
The committee reads losses, and it is entitled to. A rating built from plays will always forgive a close loss more than a résumé will.
There is also a caveat we owe you here. Between 38% and 46% of every team's rating still comes from its preseason estimate at this point in the year, and that share keeps shrinking with each game. Ohio State and Notre Dame came into the season with the first and third highest preseason ratings in the country. Notre Dame's current rating sits slightly above where it started. Ohio State's sits slightly below. So the model isn't just remembering August, but August still carries weight, and the committee gives it none.
Kentucky: the committee's No. 6, our No. 55
This is the widest gap on either list. Kentucky is 7-1 against the second-hardest schedule in the country by our measure, which is a real accomplishment and exactly what a résumé committee should reward. But our model rates the Wildcats' offense 56th and their defense 58th. A team that wins seven of eight against that schedule with middling efficiency on both sides is winning close games. Close games tend not to repeat. Kentucky plays at Tennessee on Saturday, and the model gives the Wildcats a 34% chance there.
UCLA: unbeaten, our No. 30
UCLA is 8-0 and the committee has it eighth. Our model has the Bruins 30th: the 12th-best defense in the country, but an offense rated 68th and the 74th-ranked schedule. To be fair to UCLA, it is one of the biggest risers in our ratings this season. The model started it below average and has moved it up more than almost any team in FBS. It just hasn't moved it into the top 12. UCLA is at Minnesota on Saturday, and the model has the Bruins at 70%.
LSU: our No. 3, not in the committee's top 12
LSU has the best defense in the country by our numbers. It is also the biggest riser among our top teams: its rating has climbed further above its preseason estimate than any other team in our top 12. At 6-2, the committee isn't there yet. Saturday's game with Alabama, our No. 8, settles a lot of it either way.
Oregon: our No. 7 at 5-3
Three losses usually ends a playoff conversation, and Oregon isn't in the committee's 12. The model still rates the Ducks seventh, mostly on a top-five defense and the 16th-hardest schedule. Oregon is also one of only two teams in our top 12 whose rating has fallen since August, so the model is not ignoring the losses. It thinks they mean less than three losses usually do. Oregon plays at Ohio State on Saturday.
Saturday tests both lists
Four games this weekend pit a model favourite against a committee favourite, or decide which list was closer.
| Game | Our model's win chance |
|---|---|
| Miami at Notre Dame | Notre Dame 61% |
| Oregon at Ohio State | Ohio State 72% |
| Alabama at LSU | LSU 63% |
| Kentucky at Tennessee | Tennessee 66% |
These are the model's own probabilities, before any adjustment toward the market, from the college football games board. They are not picks, and roughly one in three games like these goes the other way. The committee's second rankings come out next Tuesday night. We'll know by then which list the weekend agreed with.
How our record looks, in full
A ranking piece from a model is only worth reading if the model's own record is on the table next to it. Ours is mixed, and the weak part is as big as the strong part.
Every staked pick is booked to win one unit at the vig-free price of the side taken, so there is no sportsbook margin in any of these figures, and breakeven is the price itself, not 50%. Through Week 10:
| Market | Settled | Record | Units | Return | Early line value |
|---|---|---|---|---|---|
| Spreads | 156 | 69-87 | -15.8 | -10.3% | +0.19 pts |
| Totals | 75 | 41-34 | +8.6 | +11.8% | +0.07 pts |
| Moneylines | 144 | 28-116 | +0.9 | +1.0% | +0.59 prob pts |
| All staked picks | 375 | 138-237 | -6.3 | -2.0% | +0.15 pts |
Early line value is how far the market's line moved toward our side between when we published and kickoff. It is the measure we care about most, because it says whether we were early on information the market later agreed with.
Spreads are the problem. They have lost 15.8 units, and although the average line moved slightly our way, it moved our way on only a third of our spread picks. That is the weakest number on this page and we would rather you see it here than find it on the record page.
Totals are the bright spot at 41-34 and nearly 12% returned. The moneyline record looks alarming at 28-116, but those are mostly longshot prices where a 19% hit rate roughly breaks even, and the units say exactly that.
Two more things the record shows. First, the model also publishes leans, disagreements of four points or more where no executable contract was available. Those carry no record and no units, so the only evidence is line movement: across 388 closed leans, the line moved toward our number by 1.2 points on average. Second, we keep a shadow record of picks our own filters turned down, settled as if we had taken them. The filter that rejects edges too large to be believable has done its job, saving 6.3 units over 25 picks. The filter that rejects edges too small has cost us: those 112 declined picks went 62-50 for 11.9 units. That is a finding, and it goes back to the people who set the filters.
For calibration, the model's walk-forward backtest over 7,606 games from 2017 to 2025 shows its margin error at 12.9 points against 12.2 for the consensus closing line. That is a reference for how close to the market a model like this can get. We don't grade ourselves on it.
What happens next
The committee publishes again on Tuesday, November 10, then every Tuesday until the bracket on December 6. Our ratings update after every game. We'll come back to this comparison once the field is set, and the record above will still be on the page, whichever way it goes.
Nothing here is betting advice. If you bet, do it where it is legal, at 21 or older, and with money you can afford to lose.
College football games · Team ratings · The record · How the model works