the poll

Searching for the most accurate Top 25 in college football.

An open-source AI model rates all 138 teams and shows you every game behind the number. It also prints the scorecard where it still loses to the rating systems it is chasing.

The 2026 projection

published 20 August · 138 teams rated · never edited after the fact

The big number is projected wins on that team’s own schedule. The small one under it is the rating the board is ordered by, which is why more wins can still mean a lower rank. Tap a row for the games behind it.

Why a team with more projected wins can still rank higher.

Ohio State8.8 on its own schedule10.1 on Texas Tech’s

Texas Tech9.6 on its own schedule8.1 on Ohio State’s

Ohio State projects 8.8 wins and Texas Tech projects 9.6, and Ohio State still ranks higher. Run each on the other's calendar and the reason is plain: Ohio State would win 10.1 games on Texas Tech's schedule, while Texas Tech would win 8.1 on Ohio State's.

Built from: Built by AI using five seasons of public college football data: 8,117 games and 1,235,232 plays, with 1,000 simulated seasons to test how firm each ranking is. every file →
This view: thepoll.ai/cfb-poll
Licence: Ratings CC BY 4.0, code MIT. No account, no email. terms →
Take it: What is on screen, as an image 25 teams
Every rated team, as an image 138 teams

the same team, two calendars

Texas projects 7.1 wins on its own schedule and 8.4 on Clemson’s.

7.1on Texas’s own schedule, the hardest in the country
8.4on Clemson’s schedule, an ordinary one

Same roster, same coaches, an ordinary calendar instead of the hardest one in the country. That is what a schedule is worth.

the lookback

In week 5 of 2025 the weekly poll had Navy 22nd at 4-0. Ask it about that same week now and it says 40th.

The board at the top of this page

The projection. 138 teams ranked on 20 August, before anybody played, never edited after the fact.

A second ranking, from week 5

The poll. It starts every team at zero and reads only games that have actually been played, so it has nothing to say yet. It opens in week 5.

What week 5 said then

What week 5 says now

Navy, 4-0 at the time

the same games, worth what those opponents turned out to be

Every week the poll is built again from scratch, so a September win is worth whatever that opponent turned out to be. A rating that updates game by game cannot go back and redo September.

How far the average top-25 team ended up from where the finished season put it, week by week. Five games is thin evidence and the poll says so.

See week 5 both ways, every team →

the bar

5 bars, written down before a single result was looked at. This poll clears 0 of them.

the bar, fixed before this poll ran
this poll · FAIL
rating systems somebody else publishes
plain baselines, not rating systems
earlier versions of this model

Call more games right than anybody else does.
Straight-up accuracy at or above the floor
the bar 70.0%
68.7% FAIL
54.6%further right is better71.7%
- along the line, low to high
- the bar70.0%
Miss the final score by fewer points.
Mean absolute error at or below the ceiling
the bar 12.8
13.04 FAIL
12.5further left is better15.8
- along the line, low to high
- the bar12.8
And keep the blowout misses rare.
Root mean squared error at or below the ceiling
the bar 15.8
16.53 FAIL
15.3further left is better20.4
- along the line, low to high
- the bar15.8
When it says 70%, be right about seven times in ten.
Worst decile calibration deviation within tolerance
the bar 5.0
7.37 FAIL
4.7further left is betterno other system publishes this one, so there is nothing to compare7.7
- along the line, low to high
- the bar5.0
Put the loser above the winner less often than anybody else does.
Retrodictive violations at or below every baseline
the bar is every other system
20.2% FAIL
17.6%further left is better25.1%
- along the line, low to high

Every bar sits above the best number any system scored here puts up, and each of those systems is a named mark on the same line. Some you want high and some you want low, so the word beside the mark is the verdict, not the side it is on. A bar drawn around the systems you beat is not a bar.

Failing a bar set at somebody else’s best number does not make this ranking worse than the ones you already read. It means it is not better yet, and closing that gap is what the season is for.

2 more standards were written down and one season of data cannot settle them. They count as neither a pass nor a fail. The whole scoreboard, thresholds included →

where you argue

Tell it what it gets wrong.

You do not need code and you do not need an account. Send it to challenge@thepoll.ai and you get one of two answers: why not, argued from the same numbers, or your idea built and run in public. Both get written up on this site next to the poll.

The Story Why somebody who does not write Python had an AI build a ranking system, and what it got wrong last year.
The Model How it works, and the 7 standards it is measured against.
The Data Every game, every play, and the list of what it refuses to read.
The Invite Move a setting, send an argument, or build your own. Your name goes on what gets used.

The Poll is an open-source project. Everything it runs on is published, including the misses. A new poll goes up every Sunday of the season, posted by hand until the automation is switched on. The code is at github.com/vyhlidal/cfb-poll.

projected 1 of 138

Ohio State

This poll1
AP1
Coaches1

8.8wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 9.8.

How good the model thinks they are

1st of 13822.7 rating, and it is what the board is sorted by

How hard their schedule is

5th of 13810.0 against a field whose middle is 4.9

home games7

conference, which the model never readsBig Ten

Ohio State returns most of last year's offense, worth 1.4 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Ball State	97%	W
2	away	Texas	55%	W
3	home	Kent State	96%	W
4	home	Illinois	84%	W
5	away	Iowa	60%	W
6	home	Maryland	89%	W
7	away	Indiana	42%	L
9	away	USC	52%	W
10	home	Oregon	66%	W
11	home	Northwestern	87%	W
12	away	Nebraska	72%	W
13	home	Michigan	81%	W

Two numbers come out of that table Picking every favorite is not the same as expecting to win every one. A 75 percent favorite loses one week in four, so the wins always add up to fewer than the picks.

Projected wins8.8the 12 chances above, added up

Games it would pick11–1every game it makes the favorite, counted

The Host, on what the model found The four games Ohio State is least sure of are all on the road, and the only one the model calls a loss is at Indiana, 42 percent. Which road trip worries you most?

projected 2 of 138

Oregon

This poll2
AP2
Coaches2

9.0wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 9.6.

How good the model thinks they are

2nd of 13821.0 rating, and it is what the board is sorted by

How hard their schedule is

37th of 1387.2 against a field whose middle is 4.9

home games7

conference, which the model never readsBig Ten

Oregon returns most of last year's offense, worth 1.2 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Boise State	79%	W
2	away	Oklahoma State	88%	W
3	home	Portland State · not rated: no offseason to read	99%	W
4	away	USC	48%	L
6	home	UCLA	91%	W
7	home	Nebraska	86%	W
8	away	Illinois	61%	W
9	home	Northwestern	85%	W
10	away	Ohio State	34%	L
11	home	Michigan	78%	W
12	away	Michigan State	76%	W
13	home	Washington	75%	W

Projected wins9.0the 12 chances above, added up

Games it would pick10–2every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Only one game on Oregon's card is anywhere near a coin flip: at USC, 48 percent. Set aside the trip to Ohio State and the other ten are all 61 percent or better. Does that look right?

projected 3 of 138

Indiana

This poll3
AP6
Coaches6

9.0wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 9.5.

How good the model thinks they are

3rd of 13820.7 rating, and it is what the board is sorted by

How hard their schedule is

23rd of 1388.2 against a field whose middle is 4.9

home games8

conference, which the model never readsBig Ten

Indiana lost most of last year's offense, costing it 1.7 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	North Texas	83%	W
2	home	Howard · not rated: no offseason to read	98%	W
3	home	Western Kentucky	86%	W
4	home	Northwestern	84%	W
5	away	Rutgers	71%	W
6	away	Nebraska	68%	W
7	home	Ohio State	58%	W
8	away	Michigan	56%	W
9	home	Minnesota	85%	W
11	home	USC	71%	W
12	away	Washington	52%	W
13	home	Purdue	91%	W

Projected wins9.0the 12 chances above, added up

Games it would pick12–0every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found The model favors Indiana in all twelve games and still says 9-3, because a 75 percent favorite loses one week in four. Where do the three losses come from?

projected 4 of 138

Notre Dame

This poll4
AP4
Coaches5

9.4wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 9.5.

How good the model thinks they are

4th of 13820.5 rating, and it is what the board is sorted by

How hard their schedule is

62nd of 1385.8 against a field whose middle is 4.9

home games6

conference, which the model never readsFBS Independents

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	neutral	Wisconsin	81%	W
2	home	Rice	93%	W
3	home	Michigan State	90%	W
4	away	Purdue	77%	W
5	away	North Carolina	77%	W
6	home	Stanford	92%	W
7	away	BYU	46%	L
9	neutral	Navy	75%	W
10	home	Miami	65%	W
11	home	Boston College	92%	W
12	home	SMU	76%	W
13	away	Syracuse	78%	W

Projected wins9.4the 12 chances above, added up

Games it would pick11–1every game it makes the favorite, counted

The Host, on what the model found The only game Notre Dame is not favored in is at BYU, 46 percent, and it projects 9.4 wins on the 62nd hardest schedule of 138. Tell me which of those numbers you would move.

projected 5 of 138

Texas Tech

This poll5
AP12
Coaches12

9.6wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 9.2.

How good the model thinks they are

5th of 13819.1 rating, and it is what the board is sorted by

How hard their schedule is

83rd of 1383.9 against a field whose middle is 4.9

home games7

conference, which the model never readsBig 12

Texas Tech returns most of last year's offense, worth 0.7 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Abilene Christian · not rated: no offseason to read	94%	W
2	away	Oregon State	78%	W
3	home	Houston	79%	W
4	home	Sam Houston	94%	W
5	away	Colorado	70%	W
7	home	Arizona State	81%	W
8	away	Cincinnati	65%	W
9	home	Arizona	76%	W
10	home	West Virginia	89%	W
11	away	Oklahoma State	86%	W
12	away	Baylor	64%	W
13	home	TCU	79%	W

Projected wins9.6the 12 chances above, added up

Games it would pick12–0every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Texas Tech is favored in all twelve games and its lowest number all year is 64 percent, at Baylor. It still projects 9.6 wins, because an 80 percent favorite drops one game in five. Tell me where that is wrong.

projected 6 of 138

Miami

This poll6
AP7
Coaches7

9.2wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 9.2.

How good the model thinks they are

6th of 13819.1 rating, and it is what the board is sorted by

How hard their schedule is

74th of 1384.6 against a field whose middle is 4.9

home games7

conference, which the model never readsACC

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	away	Stanford	76%	W
2	home	Florida A&M · not rated: no offseason to read	98%	W
3	away	Wake Forest	66%	W
4	home	Central Michigan	88%	W
5	away	Clemson	59%	W
7	home	Florida State	82%	W
8	home	Pittsburgh	79%	W
9	away	North Carolina	74%	W
10	away	Notre Dame	35%	L
11	home	Duke	83%	W
12	home	Virginia Tech	90%	W
13	home	Boston College	91%	W

Projected wins9.2the 12 chances above, added up

Games it would pick11–1every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found The model gives Miami 66 percent or better in ten of its twelve games. The two that are not: at Clemson, 59, and at Notre Dame, 35. Which of the two decides their season?

projected 7 of 138

Georgia

This poll7
AP3
Coaches3

8.3wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 9.0.

How good the model thinks they are

7th of 13817.7 rating, and it is what the board is sorted by

How hard their schedule is

19th of 1388.4 against a field whose middle is 4.9

home games7

conference, which the model never readsSEC

Georgia returns most of last year's offense, worth 1.6 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Tennessee State · not rated: no offseason to read	99%	W
2	home	Western Kentucky	82%	W
3	away	Arkansas	65%	W
4	home	Oklahoma	67%	W
5	home	Vanderbilt	69%	W
6	away	Alabama	45%	L
7	home	Auburn	80%	W
9	neutral	Florida	68%	W
10	away	Ole Miss	44%	L
11	home	Missouri	73%	W
12	away	South Carolina	55%	W
13	home	Georgia Tech	79%	W

Projected wins8.3the 12 chances above, added up

Games it would pick10–2every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Georgia plays four road games and they are the four closest games on the card: Arkansas, Alabama, Ole Miss, South Carolina. The model still has them first in the SEC. Pick the one they drop.

projected 8 of 138

Texas A&M

This poll8
AP8
Coaches8

8.1wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.8.

How good the model thinks they are

8th of 13816.7 rating, and it is what the board is sorted by

How hard their schedule is

29th of 1387.9 against a field whose middle is 4.9

home games7

conference, which the model never readsSEC

Texas A&M returns most of last year's offense, worth 1.4 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Missouri State	88%	W
2	home	Arizona State	77%	W
3	home	Kentucky	85%	W
4	away	LSU	57%	W
5	home	Arkansas	83%	W
6	away	Missouri	49%	L
7	home	The Citadel · not rated: no offseason to read	97%	W
8	away	Alabama	43%	L
10	away	South Carolina	53%	W
11	home	Tennessee	69%	W
12	away	Oklahoma	41%	L
13	home	Texas	66%	W

Projected wins8.1the 12 chances above, added up

Games it would pick9–3every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Texas A&M is one of only three teams in the top 25 where the model, the AP and the Coaches all land on the same number. All three say 8. Does agreement make you more confident, or less?

projected 9 of 138

BYU

This poll9
AP14
Coaches15

8.4wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.7.

How good the model thinks they are

9th of 13816.6 rating, and it is what the board is sorted by

How hard their schedule is

45th of 1386.8 against a field whose middle is 4.9

home games7

conference, which the model never readsBig 12

BYU returns most of last year's offense, worth 2.3 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Utah Tech · not rated: no offseason to read	97%	W
2	home	Arizona	72%	W
3	away	Colorado State	78%	W
5	away	TCU	53%	W
6	home	Iowa State	81%	W
7	home	Notre Dame	54%	W
8	away	UCF	65%	W
9	home	Arizona State	77%	W
10	away	Utah	39%	L
11	home	Baylor	79%	W
12	away	Kansas	60%	W
13	home	Cincinnati	81%	W

Projected wins8.4the 12 chances above, added up

Games it would pick11–1every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found No team in the top 25 gained more from the offseason than BYU: returning production is worth 2.3 points to the projection. The model has them 9th, the AP has them 14th. Buying it?

projected 10 of 138

USC

This poll10
AP14
Coaches14

7.8wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.7.

How good the model thinks they are

10th of 13816.4 rating, and it is what the board is sorted by

How hard their schedule is

9th of 1389.6 against a field whose middle is 4.9

home games7

conference, which the model never readsBig Ten

USC returns most of last year's offense, worth 1.1 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Fresno State	83%	W
1	home	San José State	90%	W
2	home	Louisiana	84%	W
3	away	Rutgers	62%	W
4	home	Oregon	52%	W
5	home	Washington	66%	W
6	away	Penn State	51%	W
8	away	Wisconsin	64%	W
9	home	Ohio State	48%	L
11	away	Indiana	29%	L
12	home	Maryland	81%	W
13	away	UCLA	67%	W

Projected wins7.8the 12 chances above, added up

Games it would pick10–2every game it makes the favorite, counted

The Host, on what the model found Three games on USC's card read 48, 51 and 52 percent: Ohio State at home, Penn State on the road, Oregon at home. Take two of the three and where does that put them?

projected 11 of 138

Utah

This poll11
AP21
Coaches21

8.3wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.6.

How good the model thinks they are

11th of 13816.0 rating, and it is what the board is sorted by

How hard their schedule is

49th of 1386.6 against a field whose middle is 4.9

home games7

conference, which the model never readsBig 12

Utah has a new head coach, costing it 1.5 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Idaho · not rated: no offseason to read	94%	W
2	home	Arkansas	82%	W
3	home	Utah State	80%	W
4	away	Iowa State	59%	W
6	home	Kansas	80%	W
7	away	Colorado	63%	W
8	home	Houston	74%	W
9	away	Cincinnati	59%	W
10	home	BYU	61%	W
11	away	Arizona	47%	L
12	away	TCU	51%	W
13	home	West Virginia	85%	W

Projected wins8.3the 12 chances above, added up

Games it would pick11–1every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Of the top 25 teams both human polls rank, Utah is the model's biggest argument: 11th here, 21st in both. That is after 1.5 points came off for a coaching change. Who is closer?

projected 12 of 138

Oklahoma

This poll12
AP10
Coaches9

7.3wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.5.

How good the model thinks they are

12th of 13815.4 rating, and it is what the board is sorted by

How hard their schedule is

6th of 1389.9 against a field whose middle is 4.9

home games6

conference, which the model never readsSEC

Oklahoma returns most of last year's offense, worth 1.6 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	UTEP	91%	W
2	away	Michigan	44%	L
3	home	New Mexico	79%	W
4	away	Georgia	33%	L
6	neutral	Texas	51%	W
7	home	Kentucky	83%	W
8	away	Mississippi State	58%	W
9	home	South Carolina	73%	W
10	away	Florida	52%	W
11	home	Ole Miss	64%	W
12	home	Texas A&M	59%	W
13	away	Missouri	46%	L

Projected wins7.3the 12 chances above, added up

Games it would pick9–3every game it makes the favorite, counted

The Host, on what the model found Half of Oklahoma's schedule is a coin flip. Six of the twelve games sit between 40 and 60 percent, the Red River game at 51 among them. Which of the six do you take?

projected 13 of 138

Texas

This poll13
AP5
Coaches4

7.1wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.4.

How good the model thinks they are

13th of 13815.0 rating, and it is what the board is sorted by

How hard their schedule is

1st of 13811.6 against a field whose middle is 4.9

home games7

conference, which the model never readsSEC

Texas returns most of last year's offense, worth 0.7 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Texas State	74%	W
2	home	Ohio State	45%	L
3	home	UTSA	74%	W
4	away	Tennessee	42%	L
6	neutral	Oklahoma	49%	L
7	home	Florida	74%	W
8	home	Ole Miss	63%	W
9	home	Mississippi State	79%	W
10	away	Missouri	45%	L
11	away	LSU	53%	W
12	home	Arkansas	81%	W
13	away	Texas A&M	34%	L

Projected wins7.1the 12 chances above, added up

Games it would pick7–5every game it makes the favorite, counted

The Host, on what the model found Texas has the hardest schedule of all 138 and projects 7.1 wins. The same roster on a median schedule projects 8.4. The AP says 5th, the model says 13th. Which schedule should a ranking price?

projected 14 of 138

Ole Miss

This poll14
AP9
Coaches10

7.6wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.3.

How good the model thinks they are

14th of 13814.6 rating, and it is what the board is sorted by

How hard their schedule is

15th of 1388.7 against a field whose middle is 4.9

home games7

conference, which the model never readsSEC

Ole Miss has a new head coach, costing it 1.5 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	neutral	Louisville	57%	W
2	home	Charlotte	91%	W
3	home	LSU	75%	W
4	away	Florida	50%	W
6	away	Vanderbilt	38%	L
7	home	Missouri	68%	W
8	away	Texas	37%	L
9	home	Auburn	75%	W
10	home	Georgia	56%	W
11	away	Oklahoma	36%	L
12	home	Wofford · not rated: no offseason to read	94%	W
13	home	Mississippi State	78%	W

Projected wins7.6the 12 chances above, added up

Games it would pick9–3every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Ole Miss plays four road games and the model's best number in any of them is 50 percent, at Florida. The other three read 38, 37 and 36. Where does that season finish?

projected 15 of 138

Alabama

This poll15
AP13
Coaches11

7.5wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.3.

How good the model thinks they are

15th of 13814.6 rating, and it is what the board is sorted by

How hard their schedule is

17th of 1388.6 against a field whose middle is 4.9

home games7

conference, which the model never readsSEC

Alabama lost most of last year's offense, costing it 0.7 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	East Carolina	76%	W
2	away	Kentucky	61%	W
3	home	Florida State	74%	W
4	home	South Carolina	71%	W
5	away	Mississippi State	57%	W
6	home	Georgia	55%	W
7	away	Tennessee	41%	L
8	home	Texas A&M	57%	W
10	away	LSU	52%	W
11	away	Vanderbilt	38%	L
12	home	Chattanooga · not rated: no offseason to read	95%	W
13	home	Auburn	75%	W

Projected wins7.5the 12 chances above, added up

Games it would pick10–2every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Alabama and Ole Miss carry the same projected rating to the tenth, 14.6, and the board separates them by one place. They do not meet this season. Which one is actually higher?

projected 16 of 138

Vanderbilt

This poll16
APNR
CoachesNR

7.6wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.3.

How good the model thinks they are

16th of 13814.4 rating, and it is what the board is sorted by

How hard their schedule is

20th of 1388.4 against a field whose middle is 4.9

home games7

conference, which the model never readsSEC

Vanderbilt lost most of last year's offense, costing it 0.6 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Austin Peay · not rated: no offseason to read	90%	W
2	home	Delaware	82%	W
3	home	NC State	72%	W
4	away	Auburn	52%	W
5	away	Georgia	31%	L
6	home	Ole Miss	62%	W
7	home	Arkansas	80%	W
8	away	Kentucky	61%	W
10	away	Mississippi State	56%	W
11	home	Alabama	62%	W
12	away	Florida	50%	L
13	home	Tennessee	65%	W

Projected wins7.6the 12 chances above, added up

Games it would pick10–2every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Vanderbilt is 16th here and unranked by both human polls. The model puts them above Washington, Tennessee and Michigan without reading a single vote. Somebody is wrong. Who?

projected 17 of 138

Washington

This poll17
AP17
Coaches19

7.4wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.3.

How good the model thinks they are

17th of 13814.4 rating, and it is what the board is sorted by

How hard their schedule is

13th of 1389.0 against a field whose middle is 4.9

home games7

conference, which the model never readsBig Ten

Washington returns most of last year's offense, worth 0.7 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Washington State	70%	W
2	home	Utah State	78%	W
3	home	Eastern Washington · not rated: no offseason to read	94%	W
4	home	Minnesota	75%	W
5	away	USC	34%	L
6	home	Iowa	66%	W
7	away	Purdue	65%	W
9	away	Nebraska	55%	W
10	home	Penn State	70%	W
11	away	Michigan State	64%	W
12	home	Indiana	48%	L
13	away	Oregon	25%	L

Projected wins7.4the 12 chances above, added up

Games it would pick9–3every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found The only home game the model does not favor Washington in is Indiana, 48 percent. Oregon and USC both come on the road. Read the schedule column and tell me where the losses are.

projected 18 of 138

SMU

This poll18
AP19
Coaches20

8.1wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.1.

How good the model thinks they are

18th of 13813.5 rating, and it is what the board is sorted by

How hard their schedule is

66th of 1385.2 against a field whose middle is 4.9

home games7

conference, which the model never readsACC

SMU returns most of last year's offense, worth 0.7 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	away	Florida State	49%	L
2	home	UC Davis · not rated: no offseason to read	87%	W
3	away	Louisville	42%	L
4	home	Missouri State	84%	W
5	home	Boston College	85%	W
7	home	Virginia	71%	W
8	home	California	79%	W
9	away	Syracuse	65%	W
10	home	Virginia Tech	84%	W
11	home	Wake Forest	76%	W
12	away	Notre Dame	24%	L
13	away	Stanford	66%	W

Projected wins8.1the 12 chances above, added up

Games it would pick9–3every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found The model calls SMU 9-3 and puts two of the three losses in the first three weeks, both away: Florida State at 49 percent and Louisville at 42. Start 1-2 and is anybody still buying 18th?

projected 19 of 138

Tennessee

This poll19
AP20
Coaches18

7.3wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 8.0.

How good the model thinks they are

19th of 13813.1 rating, and it is what the board is sorted by

How hard their schedule is

31st of 1387.8 against a field whose middle is 4.9

home games7

conference, which the model never readsSEC

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Furman · not rated: no offseason to read	95%	W
2	away	Georgia Tech	48%	L
3	home	Kennesaw State	82%	W
4	home	Texas	58%	W
5	home	Auburn	72%	W
6	away	Arkansas	56%	W
7	home	Alabama	59%	W
8	away	South Carolina	45%	L
10	home	Kentucky	79%	W
11	away	Texas A&M	31%	L
12	home	LSU	72%	W
13	away	Vanderbilt	35%	L

Projected wins7.3the 12 chances above, added up

Games it would pick8–4every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found The model calls every Tennessee home game a win and four of the five road games losses. The whole 8-4 is about where the bus goes. Give me one they steal.

projected 20 of 138

Iowa

This poll20
AP22
Coaches22

7.4wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 7.9.

How good the model thinks they are

20th of 13812.6 rating, and it is what the board is sorted by

How hard their schedule is

42nd of 1386.9 against a field whose middle is 4.9

home games7

conference, which the model never readsBig Ten

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Northern Illinois	88%	W
2	home	Iowa State	74%	W
3	home	Northern Iowa · not rated: no offseason to read	94%	W
4	away	Michigan	38%	L
5	home	Ohio State	40%	L
6	away	Washington	34%	L
8	away	Minnesota	48%	L
9	home	Wisconsin	78%	W
10	away	Northwestern	48%	L
11	home	Purdue	82%	W
12	away	Illinois	43%	L
13	home	Nebraska	74%	W

Projected wins7.4the 12 chances above, added up

Games it would pick6–6every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Nobody in the model's top 25 is favored in fewer games than Iowa, six of twelve, and three of the other six read 43 percent or better. Is 20th too kind?

projected 21 of 138

Michigan

This poll21
AP16
Coaches16

6.9wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 7.9.

How good the model thinks they are

21st of 13812.6 rating, and it is what the board is sorted by

How hard their schedule is

2nd of 13810.6 against a field whose middle is 4.9

home games8

conference, which the model never readsBig Ten

Michigan has a new head coach, costing it 1.5 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Western Michigan	67%	W
2	home	Oklahoma	56%	W
3	home	UTEP	89%	W
4	home	Iowa	62%	W
5	away	Minnesota	48%	L
7	home	Penn State	67%	W
8	home	Indiana	44%	L
9	away	Rutgers	54%	W
10	home	Michigan State	81%	W
11	away	Oregon	22%	L
12	home	UCLA	80%	W
13	away	Ohio State	19%	L

Projected wins6.9the 12 chances above, added up

Games it would pick8–4every game it makes the favorite, counted

The Host, on what the model found Michigan takes two hits at once: 1.5 points off for a coaching change the model knows only as a change, and the second hardest schedule of 138. On a median slate this roster wins 7.9. Which hit is bigger?

projected 22 of 138

Boise State

This poll22
APNR
CoachesNR

6.8wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 7.8.

How good the model thinks they are

22nd of 13812.1 rating, and it is what the board is sorted by

How hard their schedule is

46th of 1386.7 against a field whose middle is 4.9

home games6

conference, which the model never readsPac-12

Boise State returns most of last year's offense, worth 1.5 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	away	Oregon	21%	L
2	home	Memphis	74%	W
3	home	South Dakota · not rated: no offseason to read	85%	W
4	away	Western Michigan	43%	L
5	home	Utah State	74%	W
6	away	Fresno State	54%	W
8	away	Washington State	41%	L
9	home	Texas State	69%	W
10	away	Colorado State	70%	W
11	home	Oregon State	84%	W
12	home	San Diego State	64%	W

Projected wins6.8the 11 chances above, added up

Games it would pick8–3every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Boise State is the best team the model rates outside the four power leagues, and it opens at Oregon, 21 percent, the hardest game on its card. Eleven games, not twelve. Does week one define them?

projected 23 of 138

Missouri

This poll23
AP25
Coaches25

6.9wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 7.7.

How good the model thinks they are

23rd of 13811.9 rating, and it is what the board is sorted by

How hard their schedule is

22nd of 1388.2 against a field whose middle is 4.9

home games7

conference, which the model never readsSEC

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Arkansas-Pine Bluff · not rated: no offseason to read	96%	W
2	away	Kansas	50%	L
3	home	Troy	77%	W
4	away	Mississippi State	51%	W
5	home	Florida	68%	W
6	home	Texas A&M	51%	W
7	away	Ole Miss	32%	L
9	away	Arkansas	53%	W
10	home	Texas	55%	W
11	away	Georgia	27%	L
12	home	Kentucky	77%	W
13	home	Oklahoma	54%	W

Projected wins6.9the 12 chances above, added up

Games it would pick9–3every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Six of Missouri's twelve games are priced between 50 and 55 percent. The model picks a side in all six and expects to lose nearly half of them anyway. Would you take the over on 6.9 wins?

projected 24 of 138

Arizona

This poll24
APNR
CoachesNR

7.0wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 7.7.

How good the model thinks they are

24th of 13811.8 rating, and it is what the board is sorted by

How hard their schedule is

25th of 1388.1 against a field whose middle is 4.9

home games7

conference, which the model never readsBig 12

Arizona returns most of last year's offense, worth 1.3 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	home	Northern Arizona · not rated: no offseason to read	91%	W
2	away	BYU	28%	L
3	home	Northern Illinois	87%	W
4	away	Washington State	41%	L
5	home	Cincinnati	72%	W
6	away	West Virginia	58%	W
8	home	Iowa State	72%	W
9	away	Texas Tech	24%	L
10	home	TCU	66%	W
11	home	Utah	53%	W
12	away	Kansas State	40%	L
13	home	Arizona State	68%	W

Projected wins7.0the 12 chances above, added up

Games it would pick8–4every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found Every loss the model gives Arizona is a road game, and it calls all seven home games wins. Neither human poll ranks them at all. Who is closer to right?

projected 25 of 138

Louisville

This poll25
AP24
CoachesNR

7.4wins projected on its own schedule

Run on the one shared calendar every team is also scored against, the same team projects 7.7.

How good the model thinks they are

25th of 13811.6 rating, and it is what the board is sorted by

How hard their schedule is

53rd of 1386.4 against a field whose middle is 4.9

home games6

conference, which the model never readsACC

Louisville lost most of last year's offense, costing it 0.9 points of rating.

Every game: the chance the model gives this team, and the side it picks

A missing week is an open date.

week	where	opponent	chance	call
1	neutral	Ole Miss	43%	L
2	home	Villanova · not rated: no offseason to read	84%	W
3	home	SMU	58%	W
4	home	Wake Forest	73%	W
5	away	NC State	43%	L
6	home	Florida State	69%	W
7	away	Syracuse	61%	W
9	home	Stanford	82%	W
10	away	Georgia Tech	45%	L
11	away	North Carolina	59%	W
12	home	Pittsburgh	65%	W
13	away	Kentucky	55%	W

Projected wins7.4the 12 chances above, added up

Games it would pick9–3every game it makes the favorite, counted

Schedule includes a promoted team.

The Host, on what the model found The AP has Louisville 24th and the Coaches poll leaves them off. The model lands on 25 and hands them six games between 43 and 59 percent. Ranked or not ranked?