The Poll.ai

2025 season · through week 9

Nobody voted on this.

Ranked by how hard each record was to get against the schedule that produced it. No preseason poll, no reputation, no conference name.

Switch below and watch which teams move.

Just WinALTERNATE LENS. The same games, ranked on different beliefs.
  1. 1BYU8-08-0Gap +44.71 in 44key 1.65Ranked 1. The model replayed this schedule 1,000 times, and this team finished between 4 and 71 of 136 in 90% of them.
  2. 2Indiana8-08-0Gap +37.71 in 47key 1.67Ranked 2. The model replayed this schedule 1,000 times, and this team finished between 1 and 46 of 136 in 90% of them.
  3. 3Ohio State7-07-0Gap +38.61 in 22key 1.35Ranked 3. The model replayed this schedule 1,000 times, and this team finished between 1 and 55 of 136 in 90% of them.
  4. 4Texas A&M8-08-0Gap +43.91 in 53key 1.72Ranked 4. The model replayed this schedule 1,000 times, and this team finished between 3 and 65 of 136 in 90% of them.
  5. 5Georgia Tech8-08-0Gap +48.41 in 18key 1.26Ranked 5. The model replayed this schedule 1,000 times, and this team finished between 5 and 88 of 136 in 90% of them.
  6. 6Navypoll has it 127-07-0Gap +50.81 in 5key 0.67Ranked 6. The model replayed this schedule 1,000 times, and this team finished between 2 and 98 of 136 in 90% of them.
  7. 7Alabama7-17-1Gap +12.41 in 15key 1.18Ranked 7. The model replayed this schedule 1,000 times, and this team finished between 2 and 78 of 136 in 90% of them.
  8. 8Texas Tech7-17-1Gap +5.91 in 5key 0.72Ranked 8. The model replayed this schedule 1,000 times, and this team finished between 2 and 57 of 136 in 90% of them.
  9. 9Georgia6-16-1Gap +14.01 in 6key 0.81Ranked 9. The model replayed this schedule 1,000 times, and this team finished between 5 and 102 of 136 in 90% of them.
  10. 10Ole Miss7-17-1Gap +16.81 in 6key 0.76Ranked 10. The model replayed this schedule 1,000 times, and this team finished between 8 and 101 of 136 in 90% of them.
  11. 11Louisville6-16-1Gap +12.01 in 4key 0.63Ranked 11. The model replayed this schedule 1,000 times, and this team finished between 5 and 95 of 136 in 90% of them.
  12. 12Oregon7-17-1Gap +5.51 in 5key 0.71Ranked 12. The model replayed this schedule 1,000 times, and this team finished between 2 and 51 of 136 in 90% of them.
  13. 13Houston7-17-1Gap +14.41 in 5key 0.68Ranked 13. The model replayed this schedule 1,000 times, and this team finished between 8 and 98 of 136 in 90% of them.
  14. 14Vanderbilt7-17-1Gap +5.61 in 4key 0.61Ranked 14. The model replayed this schedule 1,000 times, and this team finished between 2 and 57 of 136 in 90% of them.
  15. 15Miami6-16-1Gap +7.31 in 4key 0.60Ranked 15. The model replayed this schedule 1,000 times, and this team finished between 1 and 84 of 136 in 90% of them.
  16. 16Virginia7-17-1Gap +13.51 in 4key 0.59Ranked 16. The model replayed this schedule 1,000 times, and this team finished between 8 and 107 of 136 in 90% of them.
  17. 17North Texas7-17-1Gap +8.31 in 3key 0.49Ranked 17. The model replayed this schedule 1,000 times, and this team finished between 2 and 84 of 136 in 90% of them.
  18. 18Tulane6-16-1Gap +14.91 in 3key 0.48Ranked 18. The model replayed this schedule 1,000 times, and this team finished between 13 and 115 of 136 in 90% of them.
  19. 19James Madisonpoll has it 266-16-1Gap +8.91 in 3key 0.41Ranked 19. The model replayed this schedule 1,000 times, and this team finished between 2 and 99 of 136 in 90% of them.
  20. 20Memphis7-17-1Gap +8.51 in 3key 0.46Ranked 20. The model replayed this schedule 1,000 times, and this team finished between 2 and 86 of 136 in 90% of them.
  21. 21South Florida6-26-2Gap +7.21 in 3key 0.52Ranked 21. The model replayed this schedule 1,000 times, and this team finished between 4 and 91 of 136 in 90% of them.
  22. 22UNLVpoll has it 276-16-1Gap +19.41 in 3key 0.40Ranked 22. The model replayed this schedule 1,000 times, and this team finished between 34 and 127 of 136 in 90% of them.
  23. 23Texas6-26-2Gap +7.21 in 3key 0.46Ranked 23. The model replayed this schedule 1,000 times, and this team finished between 4 and 94 of 136 in 90% of them.
  24. 24Cincinnati7-17-1Gap +10.91 in 2key 0.38Ranked 24. The model replayed this schedule 1,000 times, and this team finished between 10 and 102 of 136 in 90% of them.
  25. 25Michiganpoll has it 186-26-2Gap +5.81 in 3key 0.48Ranked 25. The model replayed this schedule 1,000 times, and this team finished between 3 and 87 of 136 in 90% of them.

The right-hand number is how hard that season was to pull off against that exact schedule. One in 44 put BYU on top. One in 3 was enough for 25th. Below about 1 in 5 it stops separating teams, which is what the smaller key underneath is for. The bar under each name is where the model kept landing that team out of 136, every bar on one scale. A typical bar this week spans 94 places.

2025 week 9 · c1d2d35d · code b61a958 · Just Win · method cadaeb94 · 136 teams ranked

The tinted rows are the teams this lens moves at least five places, and each one carries both ranks.

Every column

The board above is the thirty-second version. Below it is everything the run published: the résumé, the power rating, the gap, the rank interval, and where the finished season puts each team.

Every figure the cards carry, for the 25 on the board, with nothing to click.
#teamrecordhow unlikely1 in this manyexpectedwins this schedule asked forbeat that bypowerrating (rank)gapplaces from the pollrangein 90% of replayshindsight
1BYU8-01 in 445.2+2.815.3 (11)+44.74 to 710.0
2Indiana8-01 in 475.4+2.622.3 (1)+37.71 to 460.0
3Ohio State7-01 in 224.7+2.321.4 (2)+38.61 to 550.0
4Texas A&M8-01 in 535.1+2.916.1 (8)+43.93 to 650.0
5Georgia Tech8-01 in 185.7+2.311.6 (27)+48.45 to 880.0
6Navy7-01 in 55.7+1.39.2 (38)+50.82 to 980.0
7Alabama7-11 in 154.6+2.417.1 (7)+12.42 to 780.0
8Texas Tech7-11 in 55.6+1.420.6 (4)+5.92 to 57-5.0
9Georgia6-11 in 64.3+1.712.3 (21)+14.05 to 102-2.0
10Ole Miss7-11 in 65.4+1.68.4 (42)+16.88 to 101+1.0
11Louisville6-11 in 44.7+1.312.8 (16)+12.05 to 95+1.0
12Oregon7-11 in 55.5+1.519.3 (5)+5.52 to 51+4.0
13Houston7-11 in 55.5+1.59.5 (36)+14.48 to 98-1.0
14Vanderbilt7-11 in 45.7+1.318.1 (6)+5.62 to 57-1.0
15Miami6-11 in 44.7+1.315.9 (9)+7.31 to 84+3.0
16Virginia7-11 in 45.7+1.39.5 (35)+13.58 to 107-1.0
17North Texas7-11 in 35.9+1.112.7 (18)+8.32 to 84+1.0
18Tulane6-11 in 34.9+1.16.0 (57)+14.913 to 1150.0
19James Madison6-11 in 35.1+0.911.6 (28)+8.92 to 99-2.0
20Memphis7-11 in 36.0+1.011.9 (25)+8.52 to 86-3.0
21South Florida6-21 in 34.9+1.112.8 (17)+7.24 to 91+2.0
22UNLV6-11 in 35.2+0.80.6 (86)+19.434 to 127-6.0
23Texas6-21 in 35.0+1.012.0 (23)+7.24 to 94-2.0
24Cincinnati7-11 in 26.2+0.88.1 (44)+10.910 to 102-7.0
25Michigan6-21 in 34.9+1.113.2 (14)+5.83 to 87+3.0
Show the full table: top 25 of 136, every columnALTERNATE LENS. The same games, ranked on different beliefs.
#teamunderline: 90% rank interval · league median width 94how unlikely1 in this many
1BYUBig 128-0·Power 15.25 (11)·Gap +44.75Ranked 1. The model replayed this schedule 1,000 times, and this team finished between 4 and 71 of 136 in 90% of them.1.6471 in 44
2IndianaBig Ten8-0·Power 22.27 (1)·Gap +37.73Ranked 2. The model replayed this schedule 1,000 times, and this team finished between 1 and 46 of 136 in 90% of them.1.6711 in 47
3Ohio StateBig Ten7-0·Power 21.41 (2)·Gap +38.59Ranked 3. The model replayed this schedule 1,000 times, and this team finished between 1 and 55 of 136 in 90% of them.1.3491 in 22
4Texas A&MSEC8-0·Power 16.11 (8)·Gap +43.89Ranked 4. The model replayed this schedule 1,000 times, and this team finished between 3 and 65 of 136 in 90% of them.1.7241 in 53
5Georgia TechACC8-0·Power 11.61 (27)·Gap +48.39Ranked 5. The model replayed this schedule 1,000 times, and this team finished between 5 and 88 of 136 in 90% of them.1.2561 in 18
6NavyAmerican Athletic7-0·Power 9.21 (38)·Gap +50.79Ranked 6. The model replayed this schedule 1,000 times, and this team finished between 2 and 98 of 136 in 90% of them.0.6731 in 5
7AlabamaSEC7-1·Power 17.11 (7)·Gap +12.42Ranked 7. The model replayed this schedule 1,000 times, and this team finished between 2 and 78 of 136 in 90% of them.1.1801 in 15
8Texas TechBig 127-1·Power 20.56 (4)·Gap +5.93·▼5Ranked 8. The model replayed this schedule 1,000 times, and this team finished between 2 and 57 of 136 in 90% of them.0.7171 in 5
9GeorgiaSEC6-1·Power 12.34 (21)·Gap +13.97·▼2Ranked 9. The model replayed this schedule 1,000 times, and this team finished between 5 and 102 of 136 in 90% of them.0.8121 in 6
10Ole MissSEC7-1·Power 8.42 (42)·Gap +16.83·▲1Ranked 10. The model replayed this schedule 1,000 times, and this team finished between 8 and 101 of 136 in 90% of them.0.7611 in 6
11LouisvilleACC6-1·Power 12.80 (16)·Gap +11.98·▲1Ranked 11. The model replayed this schedule 1,000 times, and this team finished between 5 and 95 of 136 in 90% of them.0.6341 in 4
12OregonBig Ten7-1·Power 19.27 (5)·Gap +5.51·▲4Ranked 12. The model replayed this schedule 1,000 times, and this team finished between 2 and 51 of 136 in 90% of them.0.7101 in 5
13HoustonBig 127-1·Power 9.50 (36)·Gap +14.44·▼1Ranked 13. The model replayed this schedule 1,000 times, and this team finished between 8 and 98 of 136 in 90% of them.0.6821 in 5
14VanderbiltSEC7-1·Power 18.13 (6)·Gap +5.62·▼1Ranked 14. The model replayed this schedule 1,000 times, and this team finished between 2 and 57 of 136 in 90% of them.0.6101 in 4
15MiamiACC6-1·Power 15.90 (9)·Gap +7.31·▲3Ranked 15. The model replayed this schedule 1,000 times, and this team finished between 1 and 84 of 136 in 90% of them.0.5961 in 4
16VirginiaACC7-1·Power 9.53 (35)·Gap +13.51·▼1Ranked 16. The model replayed this schedule 1,000 times, and this team finished between 8 and 107 of 136 in 90% of them.0.5881 in 4
17North TexasAmerican Athletic7-1·Power 12.66 (18)·Gap +8.29·▲1Ranked 17. The model replayed this schedule 1,000 times, and this team finished between 2 and 84 of 136 in 90% of them.0.4911 in 3
18TulaneAmerican Athletic6-1·Power 6.03 (57)·Gap +14.92Ranked 18. The model replayed this schedule 1,000 times, and this team finished between 13 and 115 of 136 in 90% of them.0.4781 in 3
19James MadisonSun Belt6-1·Power 11.56 (28)·Gap +8.93·▼2Ranked 19. The model replayed this schedule 1,000 times, and this team finished between 2 and 99 of 136 in 90% of them.0.4131 in 3
20MemphisAmerican Athletic7-1·Power 11.93 (25)·Gap +8.47·▼3Ranked 20. The model replayed this schedule 1,000 times, and this team finished between 2 and 86 of 136 in 90% of them.0.4631 in 3
21South FloridaAmerican Athletic6-2·Power 12.78 (17)·Gap +7.21·▲2Ranked 21. The model replayed this schedule 1,000 times, and this team finished between 4 and 91 of 136 in 90% of them.0.5241 in 3
22UNLVMountain West6-1·Power 0.56 (86)·Gap +19.43·▼6Ranked 22. The model replayed this schedule 1,000 times, and this team finished between 34 and 127 of 136 in 90% of them.0.4021 in 3
23TexasSEC6-2·Power 12.03 (23)·Gap +7.23·▼2Ranked 23. The model replayed this schedule 1,000 times, and this team finished between 4 and 94 of 136 in 90% of them.0.4601 in 3
24CincinnatiBig 127-1·Power 8.15 (44)·Gap +10.94·▼7Ranked 24. The model replayed this schedule 1,000 times, and this team finished between 10 and 102 of 136 in 90% of them.0.3751 in 2
25MichiganBig Ten6-2·Power 13.24 (14)·Gap +5.78·▲3Ranked 25. The model replayed this schedule 1,000 times, and this team finished between 3 and 87 of 136 in 90% of them.0.4761 in 3

What this board believes football results are for

Just Win, in its own words

If point differential pays, teams will run up the score on an opponent who cannot stop them. Everybody has seen it happen and nobody enjoyed it. This recipe takes the incentive off the table the only way a poll actually can, by refusing to look: the compression scale drops to a single point against a 7 point win premium, so margin can move a game's contribution by at most 1 point in 8 and beating someone by 70 is worth almost exactly what beating them by 1 is worth. There is nothing left to chase in the fourth quarter. The table is then ordered by the wins based resume, which asks one question and only one, what quality of team would have been expected to produce this record against this schedule. That is Colley's principle arrived at from the other direction, and the cost of it is written down below in full, because this recipe throws away real information on purpose and a reader is owed the size of what it threw away.

what it costs

  • It throws away information that is genuinely there. A 70 point win and a 1 point win are not the same event and this recipe values them within about 15 percent of each other. That is the whole objection to Colley and it lands.
  • It cannot move an unbeaten team, at all, ever. Expected wins approaches n from below, so an undefeated team has no finite root and every one of them lands on exactly the published bracket of +60, which is not a function of the schedule. The retroactive re-ranking that is this project's most differentiated product therefore cannot move an unbeaten team by a single place, in any week, in any direction. If September turns out to have been harder than it looked, this recipe cannot say so.
  • No team with a loss can ever be ranked above an unbeaten team. Not rarely. Never, by construction, in every season measured. In 2021 that forced a 13-0 Cincinnati to number one ahead of Alabama, Georgia and Michigan, a position no independent judge reached.
  • THE TIE-BREAK DOES NOT WORK AT THIS COMPRESSION, and the consequence is visible on the first four lines of the board. Under the house constants the margin aware resume separates teams sitting on the saturation bound, which is the one job it is kept for. At C = 1 that variant compresses too, so every unbeaten team saturates on BOTH columns and the order among them falls through to the last key in the sort, which is the team name. In 2023 the top four under this recipe is Florida State, Liberty, Michigan, Washington, in alphabetical order. That is not a ranking of those four teams. It is the absence of one, and the recipe has no way to produce a ranking of them because it has thrown away the only information that would distinguish them.
  • Compressing to a single point is far outside anything the tuning campaigns searched. Campaign 1's grid opened at C = 18 and campaign 2 only widened it upward. This constant is a value judgement taken to its logical end, not a fitted number, and nothing in this repository claims it predicts well.

what it changed

  • margin.beta_w7
  • margin.c1
  • publication.headline_orderingL4_resume

configs/recipes/just-win.toml

Full MeritMargin counts for exactly what it is worth. Nothing is protected.alternate lensThe House PollMargin in the engine, schedule odds in the headline. The published poll.the published poll

same evidence, different valuesEvery board on this page was fitted on the same 1,065 games, and the digest of that exact frame is ee59f839a0ed under all three. Switching lens changes the constants above and nothing else. Method digest for this one: cadaeb945e90.

Where next

  • Tune it yourself Move one setting and watch the board move.
  • The revision What this week looked like once the season ended.
  • Connectivity How much of the field was comparable, and what held it together.
  • Methodology Every constant, every baseline, and the lines that say FAIL.
run c1d2d35d · published 2026-08-15 17:18:21 UTC · code b61a958 · config bf0fa1a3…
q_ref 11.93 (Memphis) · β_w 7 · C 1 · h 4.869 · σ 16.224 · λ₁ 150 · λ₂ 2.0 · k 73.92 · w₁ 0.6230 · w₂ 0.1930
reproduce: uv run cfbpoll rank --season 2025 --through-week 9 --out out/
Close

rank 16

Virginia

7-1 · ACC. The model never reads conference names.

1 in 4how hard that record waskey 0.59
5.7wins this schedule asked for
+1.3how far past it they came

where the model can honestly pin this team, out of 136

Ranked 16. The model replayed this schedule 1,000 times, and this team finished between 8 and 107 of 136 in 90% of them.

8 to 107 · 99 places wide

power9.5, ranked 35
gap+13.5
resume23.0, ranked 16
hindsight17 · -1.0 against this rank