Two Teams Have Everything the Contenders Have. They Are Ranked 33rd and 35th.
Bringing your players back is worth about 1.2 wins a season, on top of whatever talent you had. Eight clubs this year have both. Six of them are ranked in the top fifteen. The other two went 4–8.
The question: which teams are better than their preseason ranking says?
The answer: Florida and South Carolina. Both have top-thirty talent and bring back most of last year’s production. Historically that combination wins about seven of every ten games.
Here is the surprising part. A preseason ranking is mostly a memory. It leans on last season’s record, and both of these teams went 4–8. The roster that produced that record is not the roster arriving in September.
What you get from it: two names to watch, and a reason that is not a hunch.
Experience has a price, and we can read it
Everyone agrees talent matters. The harder question is about continuity. Once you have accounted for talent, is bringing the same players back worth anything? Or is it just something coaches say in July?
So we took every team-season from 2015 through 2025, dropped the pandemic year, and fitted one plain equation to 1,297 of them. Wins, explained by two things: the talent a club recruited, and the share of last year’s production it returned.
Both matter, and the second is not close to a fluke. Go from the bottom tenth of returning production to the top tenth and you gain about a tenth of a win percentage; over twelve games that is 1.2 wins. And the talent term is already in the model, doing its own work.
The corner where seasons happen
Draw the line at top-thirty talent and top-quartile returning production and you get a corner of the chart. Clubs in it have won 69% of their games. Clubs with the same talent and less continuity won 59.5%.
That gap is 1.2 wins a year, across 389 club-seasons; the odds of it arising by chance are about one in ten thousand.
This year eight clubs sit in that corner. Six are exactly who you would expect: Ohio State, Oregon, Georgia, Texas, Texas A&M, Michigan — ranked first, second, fifth, seventh, eleventh and fifteenth.
The other two are Florida, ranked 33rd, and South Carolina, ranked 35th.
A preseason ranking is a memory wearing the clothes of a prediction.
The ProfessorWhy the rankings cannot see it
Because both went 4–8, and a ranking that starts a season has very little else to hold onto.
Look at how unsure everybody is. Across the twenty-five systems in the Massey composite, Florida is placed anywhere from 16th to 65th; South Carolina from 22nd to 71st. Those are the widest gaps among the eight. Nobody is confident. But the average of a lot of doubt still looks like a number, and that is the trap.
Run the equation on them and it gives Florida 7.9 wins and South Carolina 7.8, out of twelve. Not champions. But four wins better than last year, from teams the polls have parked in the thirties.
The reason it might not show up anyway
Now the part that argues against us, because it is real and it is large.
Florida plays six teams from the current top twenty-five. South Carolina plays five. Ole Miss, Texas, Georgia, Oklahoma, Alabama, Tennessee, Texas A&M — these are not schedules that let a good team look good.
We checked whether the model systematically over-rates the SEC, since that would sink the whole argument. It does, by about a fifth of a win, and the interval around that straddles zero. So the schedule does not break the prediction. It does mean an eight-win Florida and a six-win Florida could be the same team having different luck.
Which is the honest shape of this: we think they are better than 33rd and 35th. We are considerably less sure their records will say so.
What to take home: ask what the ranking is looking at
Every ranked list is built from something. The useful question is always what; the answer is usually “last time.”
That is fair when last time was measured on the same thing. It is much weaker when the thing has changed underneath. A roster that turned over; a staff that left; a school with a new intake. The number stays confident; the ground it stood on has moved.
So when you meet a ranking, ask what it saw and when it saw it. And then ask what has changed since, because that is where the ranking is blind, and blind in a way it will never announce.
This autumn, the answer to that question is called Florida, and it is called South Carolina.
Baseball is ranked by championship leverage: we simulate the rest of the season 60,000 times, then split those seasons by who won each of today's games. College football cannot be simulated that way, so it is ranked by how close to a coin flip the ratings make it — a lopsided game teaches you nothing. Tomorrow this block reports what happened to today's pick.
Notes & sources
Talent is the 247Sports composite of recruiting rankings as served by the CollegeFootballData API. Returning production is that API’s percentage of prior-season PPA — predicted points added — returning to the roster, which counts what players actually contributed rather than merely how many came back.
The model is ordinary least squares on 1,297 team-seasons, 2015 through 2025 with 2020 excluded for its shortened and uneven schedule. Winning percentage = 0.213 + 0.000392 × talent + 0.161 × returning production. The returning-production term carries SE 0.024 and p = 3.7 × 10⁻¹¹. R² is 0.136, which is worth stating plainly: these two things together explain about a seventh of the variation in winning, and the other six sevenths are coaching, health, schedule and luck.
The corner comparison uses the 70th percentile of talent and the 75th of returning production: 87 club-seasons meet both and won .691; 302 club-seasons have the talent without the continuity and won .595. Difference 1.2 wins over twelve games, z = 3.83, p = 0.0001.
Conference check, because it would have broken the argument: residuals by conference run from −0.038 (Pac-12) to +0.046 (independents). The SEC sits at −0.013, about a fifth of a win, with a 95% interval of [−0.042, +0.017] that includes zero. Applying it anyway gives Florida 7.9 wins and South Carolina 7.8.
2026 figures: Florida, talent 891.1 and 69% returning, Massey composite 33rd with systems ranging 16th to 65th. South Carolina, talent 862.9 and 69% returning, 35th, ranging 22nd to 71st. Both finished 4–8 in 2025.
One limit we will be held to. Returning production is measured in PPA share, which can exceed 100% or turn negative when a club loses productive players and keeps unproductive ones; the extreme tails are noisy and were clipped to [−20%, 130%] before fitting. Neither club is anywhere near those tails.
