The Poll.ai

2025 season · through week 12

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.

Full MeritALTERNATE LENS. The same games, ranked on different beliefs.
  1. 1Indiana11-011-0Gap +31.91 in 298key 2.47Ranked 1. The model replayed this schedule 1,000 times, and this team finished between 1 and 17 of 136 in 90% of them.
  2. 2Ohio State10-010-0Gap +32.71 in 77key 1.89Ranked 2. The model replayed this schedule 1,000 times, and this team finished between 1 and 22 of 136 in 90% of them.
  3. 3Texas Tech10-110-1Gap +6.41 in 20key 1.31Ranked 3. The model replayed this schedule 1,000 times, and this team finished between 1 and 19 of 136 in 90% of them.
  4. 4Oregon9-19-1Gap +8.01 in 19key 1.28Ranked 4. The model replayed this schedule 1,000 times, and this team finished between 1 and 31 of 136 in 90% of them.
  5. 5Utahpoll has it 158-28-2Gap -1.01 in 5key 0.68Ranked 5. The model replayed this schedule 1,000 times, and this team finished between 1 and 29 of 136 in 90% of them.
  6. 6Notre Damepoll has it 118-28-2Gap +1.91 in 7key 0.86Ranked 6. The model replayed this schedule 1,000 times, and this team finished between 1 and 32 of 136 in 90% of them.
  7. 7Texas A&Mpoll has it 110-010-0Gap +37.71 in 347key 2.54Ranked 7. The model replayed this schedule 1,000 times, and this team finished between 2 and 40 of 136 in 90% of them.
  8. 8BYU9-19-1Gap +15.51 in 44key 1.65Ranked 8. The model replayed this schedule 1,000 times, and this team finished between 3 and 47 of 136 in 90% of them.
  9. 9Georgia9-19-1Gap +12.41 in 33key 1.51Ranked 9. The model replayed this schedule 1,000 times, and this team finished between 3 and 46 of 136 in 90% of them.
  10. 10Miamipoll has it 188-28-2Gap +1.01 in 4key 0.61Ranked 10. The model replayed this schedule 1,000 times, and this team finished between 2 and 35 of 136 in 90% of them.
  11. 11Alabama8-28-2Gap +6.41 in 11key 1.05Ranked 11. The model replayed this schedule 1,000 times, and this team finished between 3 and 54 of 136 in 90% of them.
  12. 12Ole Misspoll has it 710-110-1Gap +16.21 in 21key 1.32Ranked 12. The model replayed this schedule 1,000 times, and this team finished between 8 and 67 of 136 in 90% of them.
  13. 13North Texas9-19-1Gap +8.41 in 5key 0.67Ranked 13. The model replayed this schedule 1,000 times, and this team finished between 5 and 55 of 136 in 90% of them.
  14. 14USC8-28-2Gap +3.31 in 5key 0.70Ranked 14. The model replayed this schedule 1,000 times, and this team finished between 3 and 44 of 136 in 90% of them.
  15. 15Oklahomapoll has it 108-28-2Gap +7.21 in 9key 0.95Ranked 15. The model replayed this schedule 1,000 times, and this team finished between 4 and 54 of 136 in 90% of them.
  16. 16James Madison9-19-1Gap +8.51 in 5key 0.66Ranked 16. The model replayed this schedule 1,000 times, and this team finished between 3 and 53 of 136 in 90% of them.
  17. 17Vanderbilt8-28-2Gap +3.81 in 4key 0.62Ranked 17. The model replayed this schedule 1,000 times, and this team finished between 3 and 46 of 136 in 90% of them.
  18. 18South Floridapoll has it 277-37-3Gap +1.11 in 2key 0.37Ranked 18. The model replayed this schedule 1,000 times, and this team finished between 4 and 53 of 136 in 90% of them.
  19. 19Washingtonpoll has it 297-37-3Gap +0.01 in 2key 0.29Ranked 19. The model replayed this schedule 1,000 times, and this team finished between 6 and 63 of 136 in 90% of them.
  20. 20Texas7-37-3Gap +6.11 in 4key 0.61Ranked 20. The model replayed this schedule 1,000 times, and this team finished between 7 and 67 of 136 in 90% of them.
  21. 21Tennesseepoll has it 287-37-3Gap -0.41 in 2key 0.29Ranked 21. The model replayed this schedule 1,000 times, and this team finished between 5 and 62 of 136 in 90% of them.
  22. 22Michiganpoll has it 178-28-2Gap +7.91 in 5key 0.70Ranked 22. The model replayed this schedule 1,000 times, and this team finished between 7 and 68 of 136 in 90% of them.
  23. 23Missouripoll has it 307-37-3Gap +2.11 in 2key 0.27Ranked 23. The model replayed this schedule 1,000 times, and this team finished between 8 and 72 of 136 in 90% of them.
  24. 24Virginia9-29-2Gap +7.21 in 4key 0.56Ranked 24. The model replayed this schedule 1,000 times, and this team finished between 9 and 70 of 136 in 90% of them.
  25. 25Illinois7-37-3Gap +7.61 in 3key 0.51Ranked 25. The model replayed this schedule 1,000 times, and this team finished between 10 and 82 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 298 put Indiana 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 73.5 places.

2025 week 12 · 1bb8a7ba · code b61a958 · Full Merit · method d21aaa06 · 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
1Indiana11-01 in 2987.2+3.828.1 (1)+31.91 to 170.0
2Ohio State10-01 in 776.8+3.227.3 (2)+32.71 to 220.0
3Texas Tech10-11 in 207.4+2.626.7 (3)+6.41 to 19-1.0
4Oregon9-11 in 196.4+2.624.2 (5)+8.01 to 31+1.0
5Utah8-21 in 56.3+1.724.8 (4)-1.01 to 290.0
6Notre Dame8-21 in 76.0+2.023.7 (6)+1.91 to 320.0
7Texas A&M10-01 in 3476.0+4.022.3 (7)+37.72 to 400.0
8BYU9-11 in 445.9+3.119.8 (12)+15.53 to 470.0
9Georgia9-11 in 335.9+3.120.6 (11)+12.43 to 460.0
10Miami8-21 in 46.5+1.522.0 (8)+1.02 to 350.0
11Alabama8-21 in 115.6+2.420.9 (9)+6.43 to 540.0
12Ole Miss10-11 in 217.3+2.716.2 (22)+16.28 to 67-3.0
13North Texas9-11 in 57.4+1.617.1 (19)+8.45 to 55-1.0
14USC8-21 in 56.3+1.720.9 (10)+3.33 to 44+1.0
15Oklahoma8-21 in 95.8+2.219.2 (14)+7.24 to 54+3.0
16James Madison9-11 in 57.5+1.517.4 (17)+8.53 to 53-2.0
17Vanderbilt8-21 in 46.5+1.519.5 (13)+3.83 to 46+1.0
18South Florida7-31 in 26.2+0.817.8 (16)+1.14 to 53+1.0
19Washington7-31 in 26.5+0.517.3 (18)+0.06 to 630.0
20Texas7-31 in 45.6+1.416.4 (21)+6.17 to 67-3.0
21Tennessee7-31 in 26.5+0.517.8 (15)-0.45 to 62-3.0
22Michigan8-21 in 56.4+1.616.4 (20)+7.97 to 68+1.0
23Missouri7-31 in 26.6+0.414.8 (24)+2.18 to 72-3.0
24Virginia9-21 in 47.6+1.414.7 (26)+7.29 to 70-1.0
25Illinois7-31 in 35.8+1.213.7 (29)+7.610 to 82+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 73.5how unlikely1 in this many
1IndianaBig Ten11-0·Power 28.10 (1)·Gap +31.90Ranked 1. The model replayed this schedule 1,000 times, and this team finished between 1 and 17 of 136 in 90% of them.2.4741 in 298
2Ohio StateBig Ten10-0·Power 27.31 (2)·Gap +32.69Ranked 2. The model replayed this schedule 1,000 times, and this team finished between 1 and 22 of 136 in 90% of them.1.8881 in 77
3Texas TechBig 1210-1·Power 26.73 (3)·Gap +6.44·▼1Ranked 3. The model replayed this schedule 1,000 times, and this team finished between 1 and 19 of 136 in 90% of them.1.3101 in 20
4OregonBig Ten9-1·Power 24.24 (5)·Gap +8.04·▲1Ranked 4. The model replayed this schedule 1,000 times, and this team finished between 1 and 31 of 136 in 90% of them.1.2771 in 19
5UtahBig 128-2·Power 24.83 (4)·Gap -1.03Ranked 5. The model replayed this schedule 1,000 times, and this team finished between 1 and 29 of 136 in 90% of them.0.6841 in 5
6Notre DameFBS Independents8-2·Power 23.73 (6)·Gap +1.92Ranked 6. The model replayed this schedule 1,000 times, and this team finished between 1 and 32 of 136 in 90% of them.0.8641 in 7
7Texas A&MSEC10-0·Power 22.32 (7)·Gap +37.68Ranked 7. The model replayed this schedule 1,000 times, and this team finished between 2 and 40 of 136 in 90% of them.2.5411 in 347
8BYUBig 129-1·Power 19.84 (12)·Gap +15.54Ranked 8. The model replayed this schedule 1,000 times, and this team finished between 3 and 47 of 136 in 90% of them.1.6471 in 44
9GeorgiaSEC9-1·Power 20.61 (11)·Gap +12.36Ranked 9. The model replayed this schedule 1,000 times, and this team finished between 3 and 46 of 136 in 90% of them.1.5141 in 33
10MiamiACC8-2·Power 21.98 (8)·Gap +1.04Ranked 10. The model replayed this schedule 1,000 times, and this team finished between 2 and 35 of 136 in 90% of them.0.6131 in 4
11AlabamaSEC8-2·Power 20.92 (9)·Gap +6.37Ranked 11. The model replayed this schedule 1,000 times, and this team finished between 3 and 54 of 136 in 90% of them.1.0531 in 11
12Ole MissSEC10-1·Power 16.20 (22)·Gap +16.19·▼3Ranked 12. The model replayed this schedule 1,000 times, and this team finished between 8 and 67 of 136 in 90% of them.1.3211 in 21
13North TexasAmerican Athletic9-1·Power 17.06 (19)·Gap +8.39·▼1Ranked 13. The model replayed this schedule 1,000 times, and this team finished between 5 and 55 of 136 in 90% of them.0.6661 in 5
14USCBig Ten8-2·Power 20.86 (10)·Gap +3.34·▲1Ranked 14. The model replayed this schedule 1,000 times, and this team finished between 3 and 44 of 136 in 90% of them.0.7031 in 5
15OklahomaSEC8-2·Power 19.23 (14)·Gap +7.17·▲3Ranked 15. The model replayed this schedule 1,000 times, and this team finished between 4 and 54 of 136 in 90% of them.0.9481 in 9
16James MadisonSun Belt9-1·Power 17.40 (17)·Gap +8.47·▼2Ranked 16. The model replayed this schedule 1,000 times, and this team finished between 3 and 53 of 136 in 90% of them.0.6601 in 5
17VanderbiltSEC8-2·Power 19.53 (13)·Gap +3.78·▲1Ranked 17. The model replayed this schedule 1,000 times, and this team finished between 3 and 46 of 136 in 90% of them.0.6201 in 4
18South FloridaAmerican Athletic7-3·Power 17.79 (16)·Gap +1.08·▲1Ranked 18. The model replayed this schedule 1,000 times, and this team finished between 4 and 53 of 136 in 90% of them.0.3671 in 2
19WashingtonBig Ten7-3·Power 17.29 (18)·Gap +0.02Ranked 19. The model replayed this schedule 1,000 times, and this team finished between 6 and 63 of 136 in 90% of them.0.2901 in 2
20TexasSEC7-3·Power 16.36 (21)·Gap +6.11·▼3Ranked 20. The model replayed this schedule 1,000 times, and this team finished between 7 and 67 of 136 in 90% of them.0.6051 in 4
21TennesseeSEC7-3·Power 17.81 (15)·Gap -0.36·▼3Ranked 21. The model replayed this schedule 1,000 times, and this team finished between 5 and 62 of 136 in 90% of them.0.2931 in 2
22MichiganBig Ten8-2·Power 16.37 (20)·Gap +7.86·▲1Ranked 22. The model replayed this schedule 1,000 times, and this team finished between 7 and 68 of 136 in 90% of them.0.7011 in 5
23MissouriSEC7-3·Power 14.83 (24)·Gap +2.14·▼3Ranked 23. The model replayed this schedule 1,000 times, and this team finished between 8 and 72 of 136 in 90% of them.0.2671 in 2
24VirginiaACC9-2·Power 14.68 (26)·Gap +7.21·▼1Ranked 24. The model replayed this schedule 1,000 times, and this team finished between 9 and 70 of 136 in 90% of them.0.5591 in 4
25IllinoisBig Ten7-3·Power 13.66 (29)·Gap +7.61·▲3Ranked 25. The model replayed this schedule 1,000 times, and this team finished between 10 and 82 of 136 in 90% of them.0.5111 in 3

What this board believes football results are for

Full Merit, in its own words

Best team wins. We measure as fairly as we can, and nothing else counts. This recipe takes the scoreboard at face value: the compression is switched off entirely, so a 70 point win is a 70 point win and not a 40 point win with the top trimmed off, and the win premium goes to 12 points, the top of the grid campaign 2 pre-registered. The table is then ordered by the margin aware resume, which asks how good these results say you are rather than how hard it was to produce them. Garbage time is still filtered and a kneel down is still worth nothing, because running the clock out is a clock decision and not merit, and charging a defense for allowing it would be measuring the scoreboard operator. Everything that happened while both teams were still playing is on the table, at full value, with no thumb anywhere.

what it costs

  • It pays to run up the score, and this recipe does not pretend otherwise. That is the entire objection to margin and this end of the axis does not answer it. If you believe a coach will keep the starters in against a beaten opponent when the poll rewards it, this is the recipe that rewards it.
  • It is a worse resume and a better predictor, measured. The headline ordering here is candidate B of the headline ordering study, which lost retrodictive violations to both alternatives in 16 of 16 season by surface cells and won forward ordering accuracy 0.6822 against 0.6626. Report 02 already ships a predictor: the Power column, on every row, every week.
  • It puts teams with losses above teams without them, hard. In 2023 candidate B ranked an 8-4 Kansas State twelfth and a 7-5 Texas A&M twenty-third, above an 11-1 James Madison and a 13-0 Liberty, because beating people badly and losing narrowly is what it rewards. demo/2023-recipes.md measures this exact combination of constants rather than quoting the study's.
  • The data does like these constants, and that cuts both ways. C uncapped with a 12 point win premium is campaign 2's lead 1: it beat the incumbent on margin MAE on the tune seasons (13.0039 against 13.0102) and again on 2024 (12.8536 against 12.9096), cleared its own pre-registered rule, and was blocked by an interaction with the accumulation window rather than by its own result. The objective it won on is margin MAE, which has no opinion about desert whatsoever.
  • Kendall's tau against the house poll on the 2023 final board is far below the 0.985 the project treats as the line between a convention and a dial. This is a dial. It is meant to be one.

what it changed

  • margin.beta_w12
  • margin.cuncapped
  • publication.headline_orderingL4_resume_margin

configs/recipes/full-merit.toml

The House PollMargin in the engine, schedule odds in the headline. The published poll.the published pollJust WinWin the game. Running the score up buys you nothing here.alternate lens

same evidence, different valuesEvery board on this page was fitted on the same 1,412 games, and the digest of that exact frame is 8584c6e3628d under all three. Switching lens changes the constants above and nothing else. Method digest for this one: d21aaa06428f.

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 1bb8a7ba · published 2026-08-15 17:14:16 UTC · code b61a958 · config bf0fa1a3…
q_ref 14.77 (Florida State) · β_w 12 · C uncapped · h 4.347 · σ 15.736 · λ₁ 150 · λ₂ 0.5 · k 70.79 · w₁ 0.4374 · w₂ 0.2179
reproduce: uv run cfbpoll rank --season 2025 --through-week 12 --out out/
Close

rank 12

Ole Miss

10-1 · SEC. The model never reads conference names.

1 in 21how hard that record waskey 1.32
7.3wins this schedule asked for
+2.7how far past it they came

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

Ranked 12. The model replayed this schedule 1,000 times, and this team finished between 8 and 67 of 136 in 90% of them.

8 to 67 · 59 places wide

power16.2, ranked 22
gap+16.2
resume32.4, ranked 7
hindsight15 · -3.0 against this rank