College Football Week 1

Massachusetts went 0-12 last season — not a game won, home or away — and their returning production for 2026 is a negative number. That happens when the players who left graded below replacement level, which is a statistician's way of saying the team is better off without some of them. It is the only negative overall figure of any team playing this weekend.

BOTH PROGRAMMES

Joe Harasymiak is in year two of a rebuild that has barely begun. The underlying numbers are the worst on this weekend's board almost across the line: −0.031 expected points per play on offence, a 33.6% success rate, 2.17 points per scoring opportunity, a 26.2% stuff rate and 2.12 line yards. Their offensive line lost the point of attack more than a quarter of the time. The defence allowed 0.262 expected points per play with 8.5% havoc, the lowest disruption rate anywhere on the card.

Our board has them 136th of 136.

Rutgers were 5-7 but 4-3 at SHI Stadium, and Greg Schiano is into his seventh year of a second spell. Athan Kaliakmanis returns after 100 total PPA, though Rutgers' returning passing production reads 10% and Dylan Lonergan arrived from Boston College. They signed the 36th-ranked recruiting class and ran an even portal cycle at fifteen in, fifteen out.

THE NUMBERS

We project Rutgers 49.2, Massachusetts 10.8 — 38.4 points on a total of 60.0. The books sit nearer thirty-one. SP+ is with us at 38.1; FPI is far softer at 21.2.

That seventeen-point spread between two respected national ratings on the same game is the most interesting thing here. It is a disagreement about whether a winless team stays winless.

THE READ

Everyone agrees Rutgers win comfortably. Nobody agrees by how much, and the honest answer is that a team returning a negative share of its production has no reliable baseline to project from. Preseason models are at their weakest exactly here, which is why the four numbers on this game span seventeen points.

Akron return 92% of last season's production. Wake Forest return 14%. That is the widest continuity gap in either direction on this weekend's board, and it is the reason our number sits below the market's on a game everyone expects Wake Forest to win comfortably.

BOTH PROGRAMMES

Joe Moorhead is in year three at Akron off 5-7, and essentially everyone is back: 87% of the passing, 102% of the rushing, 92% of the receiving, with Ben Finley returning. Their defence is better than the record suggests at 0.146 expected points allowed per play with 17.3% havoc — a higher disruption rate than Wake Forest's own.

Jake Dickert is in year two at Wake Forest off 8-4. Robby Ashford's 124 total PPA at 0.294 has gone and the returning passing figure is 23%. Twenty-four players arrived and twenty-three left, including four-star tackle Melvin Siani — who has gone to Texas — and defensive lineman Mateen Ibirogba.

THE NUMBERS

Wake Forest's profile is boom-or-bust: a 38.0% offensive success rate, which is poor, against 1.48 explosiveness, which is the second-highest figure of any team playing this weekend. They did not sustain drives; they hit long plays. Against an Akron defence allowing 1.31 explosiveness, that is the specific tension in the game.

Neither offence is efficient. Wake Forest managed 3.23 points per scoring opportunity, Akron 3.25 — nearly identical, and both low.

We project Wake Forest 30.7, Akron 10.4 — 20.3 points on a total of 41.1. The books are near twenty-three, SP+ 28.8, FPI 22.7. We are the low number on the spread and well under everyone on the total, which sits around forty-nine in the market.

THE READ

Our engine sees two teams that struggle to move the ball methodically and projects a low-scoring game. That is a defensible read of the success rates. The eight-point gap on the total is the claim to weigh: it requires Wake Forest's explosiveness — their one genuine strength — not to convert into points.

Eastern Michigan are 117th on our board at −9.1. San Jose State are 115th at −9.0. In a 136-team ranking built to separate teams, these two are as close to indistinguishable as the model gets — and then the roster numbers pull them apart completely.

BOTH PROGRAMMES

Eastern Michigan return 72% of last season's production and 97% of the passing. Noah Kim is back after 103 total PPA. Chris Creighton is in his thirteenth season in Ypsilanti, one of the longest tenures in the sport and a genuine rarity at this level. Twelve players arrived through the portal and twenty-six left, but the core stayed.

San Jose State return 4%. Walker Eget generated 130 total PPA at 0.312 and is gone; their returning passing figure is marginally below zero. Ken Niumatalolo is in year two after a 3-9 season with a stark split — 3-3 at home, 0-6 on the road. They have not won away from San Jose in a calendar year, and they open in Michigan.

THE NUMBERS

San Jose State's 2025 offence actually graded better — 0.172 expected points per play against Eastern Michigan's 0.199 is close, and their line yards at 3.11 are respectable — but they converted poorly, 2.77 points per scoring opportunity, among the lowest figures on the board.

Eastern Michigan's defence allowed 0.250 expected points per play, San Jose State's 0.204 with 12.1% havoc. Neither is good.

We project Eastern Michigan 28.6, San Jose State 28.0 — a 0.6-point game, essentially a coin flip, on a total of 56.6. The books have Eastern Michigan nearer four and a half, SP+ 2.1, FPI 0.5.

THE READ

All four numbers agree this is close and disagree about which shade of close. Ours is the most extreme, projecting a game inside a point. The continuity gap is the argument for the home side — one team is essentially last year's team and the other is not — and neither has shown it can win on the road. One of them has to.

Georgia Tech return 13% of last season's production. Colorado return 14%. Between them that is the emptiest pair of cupboards on the Week 1 board, and the game is essentially a test of which coaching staff restocked better.

BOTH PROGRAMMES

Brent Key is in year four at Georgia Tech off a 9-3 season that was built evenly — 5-1 at home, 4-1 away. The problem is what left. Haynes King generated 158 total PPA at 0.321 a play and is gone; Georgia Tech's returning passing production is zero percent. They brought in running back Justice Haynes, edge Noah Carter and receiver Jaylen Mbakwe, all four-star, in a nineteen-in, twenty-out portal cycle, and signed the 41st-ranked class.

Deion Sanders is in year four at Colorado off 3-9, winless on the road. Their portal traffic is among the heaviest in the sport — 43 in, 41 out — and the outgoing list is the concerning half: five-star offensive tackle Jordan Seaton left for LSU, along with four-star receiver Omarion Miller and corner DJ McKinney. In came receiver DeAndre Moore Jr., safety Boo Carter and linebacker Liona Lefau, all four-star.

Colorado's returning passing and rushing production are both negative figures, which happens when the departed players graded below replacement. It is a statistical curiosity rather than a verdict, but it does mean there is very little to project forward.

THE NUMBERS

Georgia Tech's offence was the better unit by a distance: 0.272 expected points per play on a 49.1% success rate, with a 16.1% stuff rate and 3.36 line yards — an offensive line that won the point of attack. Colorado's offence managed 0.084 on a 34.9% success rate with a 22.5% stuff rate.

Neither defence is good. Georgia Tech allowed 0.194 expected points per play with 13.7% havoc; Colorado allowed 0.190 with 15.1%. Both are in the bottom third of this weekend's units.

We project Georgia Tech 36.8, Colorado 19.8 — 17.0 points on a total of 56.6. The books have Georgia Tech by around a touchdown. That is a ten-point gap and one of our largest of the week.

THE READ

Our engine is reading two things the market is discounting: the size of the efficiency gap between the two 2025 offences, and Colorado's inability to win away from Boulder. What it is probably underweighting is that neither of those 2025 offences still exists in any meaningful sense. When both teams return roughly an eighth of their production, last season's efficiency is a weak guide, and the market's tighter number reflects that uncertainty more honestly than our confident one does.

UAB return 80% of last season's production and 100% of the passing. Illinois return 21%. On continuity alone the visitors have the better of this, which is not how a 27-point line reads.

BOTH PROGRAMMES

Bret Bielema is in year six at Illinois off 8-4, built almost entirely at home: 6-1 at Memorial Stadium, 2-3 away. Luke Altmyer generated 160 total PPA at 0.388 a play in 2025, one of the more efficient quarterback seasons in this fixture set.

Here our sources disagree with each other. Altmyer is recorded as a returning starter, while the returning-passing figure for Illinois reads 3%. Those cannot both be true, and we are not going to resolve it by choosing the more convenient one. What is certain is that 27 players left through the portal against 20 arrivals, including four-star defensive lineman Tomiwa Durojaiye and tight end Cole Rusk, and that quarterback Katin Houser arrived from East Carolina.

Alex Mortensen takes over at UAB, promoted from offensive coordinator rather than hired from outside. He inherits continuity — Jalen Kitna and Ryder Burton both return — and a defensive problem: 0.323 expected points allowed per play, the weakest defensive figure of any team playing this weekend, on a 48.8% defensive success rate and 12.0% havoc.

THE NUMBERS

Illinois' profile is solid rather than spectacular: 0.207 expected points per play on a 47.5% success rate, with 15.3% defensive havoc. UAB's offence is respectable — 0.178 and 3.11 line yards — which is why our total is so high.

We project Illinois 47.9, UAB 19.5 — 28.4 points on a total of 67.4. The books are close to us on the spread at around twenty-seven and a half, SP+ 30.4, FPI 24.3. Our total is roughly ten points above the market's.

THE READ

Rare agreement on the margin and a wide gap on the total. Our engine expects UAB's defence to leak badly enough that the scoreboard keeps moving regardless of how one-sided the game becomes — a reasonable read given that 0.323 figure, and the one number in this fixture nobody disputes.

Alabama open at home as a four-touchdown favourite and our model does not come close to the price. The gap is the largest we have with the books all weekend, and the reason is a defensive profile the market appears to be ignoring.

BOTH PROGRAMMES

Kalen DeBoer is in year three off 10-3, and the quarterback room is genuinely unsettled. Ty Simpson generated 178 total PPA at 0.322 a play and has gone. Alabama return 26% of production and 12% of the passing. Keelon Russell is the projected starter as a redshirt freshman, but our source records the competition with Austin Mack running into fall camp, so the job is not settled.

Alabama signed the second-ranked recruiting class in the country. Seventeen players arrived and twenty-four left, including four-star receiver Isaiah Horton, defensive lineman James Smith and edge Qua Russaw. This is not a settled roster asked to do a routine thing; it is a rebuilt one asked to cover four touchdowns in week one.

East Carolina are why our number is low. Blake Harrell is in year two off 8-4, 5-1 at home. Their defence graded 0.050 expected points allowed per play with 19.1% havoc — a higher havoc rate than Alabama's own 17.7%, and better in expected points than Alabama's 0.071. Their offensive line allowed a 14.3% stuff rate, the second-best figure in this fixture.

They have their own quarterback problem: Katin Houser's 149 total PPA is gone and their returning passing production is zero.

THE NUMBERS

So both teams return essentially nothing at quarterback, and one of them is being asked to give four touchdowns.

Alabama's edge is scoring efficiency — 4.04 points per scoring opportunity against East Carolina's 3.76 — and raw talent our engine cannot see in a box score. East Carolina's edge is disruption and a line that does not get stuffed.

We project Alabama 31.0, East Carolina 21.0 — 10 points on a total of 52.0. The books are up around twenty-eight and a half. SP+ says 22.4, FPI 23.2. Every independent rating sits with the market and against us.

THE READ

When our model is alone by eighteen points and three other ratings cluster on the other side, the honest reading is that our engine is over-crediting East Carolina's defensive numbers, which were produced against American Athletic offences rather than SEC ones. The havoc rate is real. Whether it survives contact with Alabama's line is the question our model does not ask and the market has already answered.

AP have Alabama 13th; our board has them 8th, five places higher than the voters.

Auburn and Baylor both went 5-7 last season, both 3-4 at home and 2-3 away — identical records in identical shapes. They meet at a neutral site in Atlanta with new quarterbacks acquired by very different routes.

BOTH PROGRAMMES

Alex Golesh takes over at Auburn from South Florida, where he went 23-15 across three seasons after a spell as Tennessee's offensive coordinator. He has brought Byrum Brown with him from USF. Auburn return 14% of production and zero percent of the passing: Jackson Arnold and Ashton Daniels are both gone, and the portal cost them five-star receiver Cam Coleman, who is now at Texas, along with four-star receiver Eric Singleton Jr. Thirty-eight arrived, thirty-nine left.

Baylor's rebuild is quieter but the acquisition is louder. Dave Aranda is in year seven, they return 60% of production and 78% of the passing with Sawyer Robertson back after 125 total PPA, and they have added DJ Lagway — the four-star quarterback whose era at Florida has just ended. That gives Baylor an unusual thing for a week-one team: an established starter and a highly rated alternative.

THE NUMBERS

Baylor's weakness is disruption: 12.3% havoc, among the lowest rates on this weekend's board, on a defence allowing 0.173 expected points per play. Auburn's own defensive figures are not in our advanced set, but the offensive rebuild is the more consequential unknown.

Baylor's offensive line graded well — a 16.2% stuff rate and 3.09 line yards.

We project Auburn 35.1, Baylor 23.7 — 11.4 points on a total of 58.8. The books have Auburn by around seven and a half. We are roughly four points firmer on the favourite and our total is close to theirs.

THE READ

Our engine is crediting Auburn's SEC talent base and discounting the fact that almost none of last season's production remains. Baylor return four times as much of themselves and have more certainty at quarterback. The market's tighter number reflects that; ours reflects the recruiting gap.

Neither team is ranked. Our board has Auburn 26th and Baylor 47th.

Brendan Sorsby produced 176 total PPA at 0.446 a play for Cincinnati last season — one of the most efficient quarterback seasons of any team playing this weekend. He transferred to Texas Tech as a five-star portal move, and Cincinnati's replacement is recorded in our data as unresolved. That single fact is the whole game.

BOTH PROGRAMMES

Scott Satterfield is in year four off 7-5, 5-2 at home. On paper Cincinnati's 2025 offence was excellent: 0.299 expected points per play and a 46.8% success rate, the best raw offensive efficiency of any team on this weekend's board. Underneath it, they return 8% of production and 4% of the passing. The number that made them look good and the players who generated it have almost entirely separated.

Boston College went 2-10 under Bill O'Brien, in year three, with a single home win. They return 16% of production and 8% of the passing, and their rushing figure is deeply negative — the signature of a run game that actively cost them points. Their defence allowed 0.235 expected points per play with 12.3% havoc, and gave up explosive plays at 1.37.

Our board separates these two teams by two places and 0.2 points of rating — 81st and 83rd. There is very little between them.

THE NUMBERS

We project Cincinnati 38.1, Boston College 22.0 — 16.1 points on a total of 60.1. The books sit at seven and a half, and both SP+ (8.0) and FPI (9.6) agree with them. We are roughly eight points more bullish on Cincinnati than anyone else.

THE READ

This is the clearest case on the card of a model reading a number that a roster change has already invalidated. Our engine is projecting Cincinnati's 2025 offensive efficiency forward onto a team that returns four percent of its passing production and has no confirmed quarterback. The market has discounted for exactly that. When three independent numbers cluster eight points below ours and the mechanism for our error is this legible, it is our number that should move.

The Border War is the oldest rivalry in the Mountain West and it opens Jim Mora's tenure at Colorado State. Our model has the visitors, which puts us on the opposite side of this game from the market, SP+ and FPI all at once.

BOTH PROGRAMMES

Colorado State went 2-10 and did not win a road game. Mora arrives from UConn, where he had rebuilt a programme from a similar floor. He inherits a heavy portal cycle — 32 in, 38 out — including quarterback Hauss Hejny from Oklahoma State, and four-star linebacker Owen Long departing. They return 40% of production but only 23% of the passing.

Wyoming were 4-8 under Jay Sawvel, in year three, and are 111th on our board — below Colorado State on rating. What our model likes is the defence: 0.069 expected points allowed per play against Colorado State's 0.237, one of the widest defensive gaps between two teams with similarly poor records anywhere on this card.

Wyoming's problem is the other half. Their offence produced 0.049 expected points per play on a 22.3% stuff rate and 2.90 points per scoring opportunity, and their returning passing production is 3% — Kaden Anderson's 32 total PPA has gone.

THE NUMBERS

We project Wyoming to win by 3.4, 20.5 to 17.1, on a total of 37.6 — the lowest total we project on any game this weekend. The books have Colorado State favoured by around three and a half with a total near forty-eight. SP+ (1.7) and FPI (3.2) both side with the market on the favourite.

THE READ

A seven-point swing on the side and more than ten on the total. Our engine is projecting a low-scoring defensive game between two 2025 also-rans, driven almost entirely by Wyoming's defensive numbers; everyone else expects Mora's arrival to lift the home side.

Rivalry openers with a first-year coach are the hardest games on any card to model, and this is the only one this weekend where our engine takes a side against all three outside ratings.

Tulane went 11-2 and did not lose in New Orleans. They arrive in Durham as clear underdogs, and both teams share the same problem: the quarterback who defined their season is playing somewhere else this year.

BOTH PROGRAMMES

Jake Retzlaff produced 150 total PPA at 0.301 for Tulane and is gone; their returning passing production is zero percent, on 21% overall. Will Hall was promoted from the offensive staff after the 11-2 season cost Tulane its head coach — an internal hire, which our engine treats more gently than an outside one, though it does not treat it as nothing.

Duke's departure is more conspicuous. Darian Mensah generated 184 total PPA at 0.343, the second-best quarterback season in this fixture set, and he has transferred to Miami — where, as it happens, he starts against Stanford this same weekend. Duke return 17% of production and 4% of the passing. Manny Diaz is in year three off 8-5, better away (4-2) than at home (3-3).

Both sides therefore have an intact idea and no one to execute it.

THE NUMBERS

Duke's offence graded higher than Tulane's — 0.245 expected points per play against 0.205, with 1.33 explosiveness and 4.32 points per scoring opportunity. Tulane's edge is the offensive line: a 17.8% stuff rate and 3.10 line yards, both better than Duke's 21.8% and 2.83.

We project Duke 37.2, Tulane 33.9 — 3.3 points on a total of 71.1. The books have Duke by nine and a half. FPI is close to us at 3.8; SP+ is the outlier the other way at 12.4.

THE READ

The spread disagreement is modest and defensible. The total is not: 71.1 against a market near fifty-two is our largest total-side claim of the weekend, and it rests on two offences whose primary producers have both left. Projecting a shootout between two teams that must replace their quarterbacks is the sort of thing a model does when it reads efficiency and ignores personnel.

Florida went 4-8 and the split is the whole story: 4-2 in Gainesville, 0-5 everywhere else. Jon Sumrall arrives from Tulane, where he had just gone 11-2 — a coach leaving a programme that won eleven for one that won four.

BOTH PROGRAMMES

The quarterback churn is worth following. DJ Lagway has left Florida for Baylor, where he opens this same weekend; Aaron Philo arrives to replace him. Florida return 69% of production and 94% of the passing, which reflects RJ Johnson III's 80 total PPA at 0.313 rather than Lagway's 51 at 0.131. Thirty-six players left through the portal against twenty-seven arrivals — including four-star receiver Eric Singleton Jr. from Auburn — and four-star edge Jayden Woods and receiver Eugene Wilson III departed.

They signed the 17th-ranked recruiting class. They signed the 17th-ranked recruiting class. Our board has Florida 22nd at +9.7 despite the record, which is the model saying the roster is better than the results. AP do not rank them.

Florida Atlantic went 4-8 under Tom Herman, in year two, with the same home-heavy shape at 3-3 and 1-5. They return more than Florida do — 68% overall, 93% passing, and a rushing figure above 100% — with Caden Veltkamp back after 114 total PPA. Their offence is boom-or-bust: 1.40 explosiveness, one of the higher figures on the board, on a modest 40.3% success rate.

THE NUMBERS

Florida Atlantic's defence is the vulnerability at 0.224 expected points allowed per play with 13.9% havoc. Florida's own defence allowed 0.172 with 14.5% — better, not dominant.

We project Florida 37.4, Florida Atlantic 22.7 — 14.7 points on a total of 60.1. The books are up near twenty-six and a half, SP+ 23.8, FPI 27.3. We are eleven points below the market and the lowest number on the game by a distance.

THE READ

Our engine prices Florida on last season's results; the market prices them on the roster and the new staff. Since our model docks new head coaches by default and Sumrall is exactly the kind of hire that default is worst at handling, this is a case where our own methodology is probably working against us.

Houston went 9-3 last season and won every road game they played — 6-0 away, 3-3 at TDECU. It is one of the strangest home-road splits in the sport, and this week they are at home.

BOTH PROGRAMMES

Willie Fritz is in year three with Conner Weigman returning after 131 total PPA at 0.284. Houston return 81% of last season's production, 92% of the passing — one of the highest continuity figures of any ranked team playing this weekend — and signed the 38th-ranked class. AP have them 23rd; our board is higher at 31st of 136 with 7.6 projected wins.

Their efficiency profile is oddly flat: 0.106 expected points per play on offence and 0.106 allowed on defence, with a 20.4% stuff rate and 2.62 line yards. That is the shape of a team that wins close games rather than dominating them, and it is worth holding against the 9-3 record.

Oregon State were 2-10 and winless on the road. Our data carries no verified head coach entry for them, so none is named here. They return 58% of production and 90% of the passing — more continuity than Houston's record would suggest they are facing — with Maalik Murphy back after 54 total PPA. Twenty arrived through the portal against thirty-five departures.

Their defence is the respectable half at 0.183 expected points allowed with 17.2% havoc; the offence produced 0.062 and 2.89 points per scoring opportunity, among the weakest on the card.

THE NUMBERS

We project Houston 37.1, Oregon State 12.3 — 24.8 points on a total of 49.4. The books sit lower at around twenty and a half, with SP+ 15.5 and FPI 17.7. We are the high number by four points against the market and nine against SP+.

THE READ

Our model is treating Oregon State's offensive collapse as durable and Houston's poor home form as noise. The market is doing close to the reverse, and it has a point: Houston's home record last season was the weaker half of a 9-3 campaign, and Oregon State return meaningfully more of themselves than a 2-10 team usually does.

North Texas went 11-2 last season and return two percent of the production that did it. Two. Not two of ten categories — two percent overall, two percent passing, three percent rushing, two percent receiving. It is the most complete roster turnover of any team playing this weekend, and it is the single most important fact in this game.

BOTH PROGRAMMES

The reason is legible. Drew Mestemaker threw for 162 total PPA at 0.366 a play and was a five-star portal departure — to Oklahoma State, following his head coach Eric Morris. Neal Brown arrives from a spell as a Texas assistant after his time as West Virginia's head coach, inheriting a 49-in, 38-out portal cycle and almost nothing else.

What North Texas leave behind on paper is spectacular: 0.351 expected points per play, a 51.6% success rate — the highest of any offence playing this weekend — 3.47 line yards and 4.98 points per scoring opportunity. None of the players who produced it are there.

Indiana went 13-0. Unbeaten, 7-0 at home and 5-0 away, and they also lost their quarterback: Fernando Mendoza's 198 total PPA at 0.468 a play is the best single quarterback season in this entire fixture set, and he is gone. Josh Hoover arrives having led the Big 12 in yards per attempt. Indiana return 17% of production and 2% of the passing, and signed the 29th-ranked class.

THE NUMBERS

Indiana's defence is the best unit on this weekend's board by a distance: 0.014 expected points allowed per play, a 34.4% defensive success rate and 23.2% havoc, the highest disruption rate anywhere on the card. Their offence graded 5.13 points per scoring opportunity, also the best figure here.

We project Indiana 44.2, North Texas 25.3 — 18.9 points on a total of 69.5. The books are up around forty-one. SP+ says 39.9, FPI 32.0. We are more than twenty points below the market and the lowest of the four numbers by a wide margin.

THE READ

This is our largest disagreement of the week, and it rests on a single assumption: that North Texas' 2025 offensive efficiency carries forward. Given they return two percent of it, that assumption is doing an enormous amount of work and almost certainly should not be. Our board also ranks Indiana 21st against AP's 6th, so we are consistently the low opinion on the undefeated team — consistency that makes the error systematic rather than random.

Kirk Ferentz begins his twenty-eighth season at Iowa, which is longer than most of the players on this field have been alive. Northern Illinois arrive under a first-year head coach in the year the school moves conferences.

BOTH PROGRAMMES

Iowa went 8-4 and AP have them 22nd; our board is lower at 38th. The defence is the identity as always — 0.055 expected points allowed per play on a 38.6% defensive success rate with 16.9% havoc — and the offence exists to not lose the game.

The quarterback position is genuinely open. Mark Gronowski's 102 total PPA at 0.289 has gone, Iowa return 16% of their passing production, and our data lists Jeremy Hecklinski as a projected starter while noting Hank Brown is also in the mix. Neither has started. Iowa's portal cycle was the quietest of any Big Ten team playing this weekend: fifteen in, eight out.

Rob Harley steps up at Northern Illinois after a spell as interim, an ex-Ohio State safety inheriting a 3-9 team mid-move to the Mountain West. They lost 22 players and brought in seven — the smallest incoming class on this weekend's board — though they do return 44% of production and a passing figure above 100%.

THE NUMBERS

Northern Illinois' defence allows 0.186 expected points per play with 11.2% havoc, the lowest disruption rate of any team playing this weekend. Their offence produced 3.04 points per scoring opportunity.

We project Iowa 35.0 and Northern Illinois to a negative figure, giving a margin of 36.5 on a total of 33.5. The books are near thirty-one and a half, SP+ 34.3, FPI 27.6.

THE READ

The negative visitor projection is a statement about the model's limits, not a forecast — a team cannot score below zero, and any engine that outputs one is extrapolating past what its inputs support. We publish it flagged rather than quietly clipping it.

On the spread, all four numbers sit inside a nine-point band. The total is where we separate: 33.5 against a market near forty-five and a half, which is our engine projecting an Iowa game in the most Iowa way imaginable.

James Madison went 12-1 last season and did not lose at home. They return eight percent of the production that did it, and the man who takes charge of the rebuild was fired mid-season by Florida.

BOTH PROGRAMMES

Billy Napier's arrival is one of the stranger moves of the cycle: a Power Four head coach dismissed in-season taking over a Sun Belt programme that had just had a far better year than the one he left. He inherits a roster stripped almost bare — 8% overall returning, 9% passing, 7% rushing — with 37 portal arrivals against 30 departures doing the restocking.

What James Madison do have on paper is a superb 2025 profile: 0.225 expected points per play on offence with 3.24 line yards, and a defence that allowed 0.002 expected points per play on a 30.1% defensive success rate with 19.9% havoc. Those defensive numbers are among the best of any team playing this weekend, Power Four included.

Liberty were 4-8 under Jamey Chadwell in year four but return more — 38% overall and 75% of the passing, with Ethan Vasko back. Our board still projects Liberty to 8.0 wins against James Madison's 7.3, which is a schedule judgement rather than a quality one.

THE NUMBERS

Liberty's defence is the weak unit: 0.204 expected points allowed per play with 13.2% havoc, the lowest disruption rate in this fixture.

We project James Madison 42.5, Liberty 18.3 — 24.2 points on a total of 60.8. The books have it at six and a half. SP+ says 6.9, FPI 8.2. Our number is roughly eighteen points clear of every other rating on the game, our second-largest disagreement of the week.

THE READ

The defensive numbers driving our projection are real, but they were produced by players who have overwhelmingly left, against Sun Belt offences, under a different coaching staff. Our engine is applying all of that forward at close to full weight. Three independent ratings say six to eight points. When the gap is eighteen and the roster returning is eight percent, the burden is entirely on us.

Lane Kiffin walks into Tiger Stadium as LSU's head coach having spent five years at Ole Miss going 55-19 with three consecutive ten-win seasons. Across from him is Dabo Swinney, entering his eighteenth year at Clemson. It is the least stable and most stable job in the sport, meeting in week one, in the loudest building in college football.

BOTH PROGRAMMES

Both went 7-5 last season and both did it in opposite shapes. LSU were 6-1 at home and 1-4 on the road — a team that could not win away from Baton Rouge. Clemson were the mirror image, 3-4 at home and 4-1 away.

What separates them now is how each rebuilt. LSU have been through the transfer portal like a weather event: 41 players in and 41 out, headlined by five-star arrivals in offensive tackle Jordan Seaton and quarterback Sam Leavitt, with edge rusher Princewill Umanmielen pulled across from Kiffin's old team at Ole Miss. They return just 21% of last season's production. Clemson barely moved — 11 in, 15 out — and return 41%.

That is two entirely different bets. LSU have bought a roster; Clemson have kept one.

THE QUARTERBACKS

Garrett Nussmeier's 2025 produced 66 total PPA at 0.211 a play. He is gone, and Leavitt arrives to run Kiffin's offence. Clemson hand the job to Chris Vizzina, who has spent three years behind Cade Klubnik and gets his first extended run; Klubnik's 98 total PPA leaves with him, which is why Clemson return only 14% of their passing production.

Neither side knows what it has at the position. Both think they have upgraded.

THE NUMBERS

Clemson's defence is the best unit on the field: 0.095 expected points allowed per play, a 37.3% defensive success rate and 19.6% havoc, the highest of the four units here. LSU's offence was the weakest — 0.087 expected points per play, a 24.2% stuff rate (the rate at which runs are stopped at or behind the line), and 2.51 line yards. That is an offensive line that lost the point of attack repeatedly.

We project LSU 19.5, Clemson 15.9 — a 3.6-point margin on a total of 35.4. The books have LSU by nine and a half with a total in the low fifties. We are six points off on the side and roughly fifteen under on the total.

THE READ

That total is the largest single disagreement on the Week 1 card, and it is ours to justify, not the market's. Our engine is projecting LSU's 2025 offensive inefficiency forward onto a roster that has replaced most of it, and pairing it with Clemson's genuine defensive quality. Both those inputs are real. The question is whether 41 incoming transfers and a new coaching staff invalidate the first one — and preseason models are worst precisely here, at the seam between last season's numbers and this season's roster.

AP have LSU 11th. Our board has them 17th, and Clemson 23rd despite no AP ranking at all.

Charles Huff arrives at Memphis from Southern Miss and has rebuilt the roster almost from scratch: 51 players in, 47 out, and 6% of last season's production returning. Two percent of the passing, zero percent of the rushing. Memphis went 8-4; almost nobody who did it is still there.

BOTH PROGRAMMES

Brendon Lewis generated 116 total PPA at 0.256 and has gone. Marcus Stokes arrived at quarterback among the intake, along with running back Dallan Hayden and safety Ian Foster. Memphis' 2025 offence graded well — 0.245 expected points per play, a 44.9% success rate and 1.31 explosiveness — and our board still has them 68th with 7.8 projected wins, the highest of any team in this fixture.

Arkansas State were 6-6 under Butch Jones, now in year seven, and their profile is flat in both directions: 0.090 expected points per play on offence, 0.192 allowed. They return 30% of production, and their rushing figure is deeply negative — the mark of a run game that cost them points. Twenty-five arrived, thirty-seven left.

THE NUMBERS

Arkansas State's offensive line is the weak point: a 21.4% stuff rate on 2.57 line yards. Their defence allows explosive plays at 1.29.

We project Memphis 42.6, Arkansas State 25.0 — 17.6 points on a total of 67.6. The books are around ten, SP+ 14.4, FPI 9.7. We are the highest number on the game by three points over SP+ and nearly eight over the market.

THE READ

Our engine rates Memphis' 2025 offensive efficiency highly and applies little discount for a coaching change that has replaced almost the entire roster. That is the same mechanism that makes our Cincinnati and North Texas numbers extreme this week: efficiency read forward, personnel largely ignored. The total is the bigger claim — 67.6 against a market near fifty-five and a half.

Kyle Whittingham spent twenty-one seasons as Utah's head coach. He opens 2026 in the Big House in different colours, which remains one of the strangest sentences of this college football year. Across from him is a Western Michigan side that returns one hundred percent of last season's production — passing, rushing and receiving, all of it.

BOTH PROGRAMMES

That Western Michigan number is not a rounding artefact. Lance Taylor is in year three off 9-4, their portal traffic was fourteen in and thirteen out, and every meaningful contributor is back. Broc Lowry returns after 80 total PPA. Their defence graded 0.053 expected points allowed per play with 17.3% havoc — better in expected points than Michigan's own 0.113.

Michigan return 69% of production and, unusually, more passing production than they had — 104%, which happens when the returning quarterback's share grows. Bryce Underwood is back and confirmed by Whittingham as his clear number one, after 100 total PPA at 0.253 as a former five-star recruit. Michigan signed the 12th-ranked recruiting class and lost 33 players to the portal against 17 in, including running back Justice Haynes to Georgia Tech.

THE NUMBERS

Michigan's offensive line is the strength: a 13.7% stuff rate, the best figure of any team playing this weekend, and 3.32 line yards. Western Michigan's offence is the limiter — 1.05 explosiveness, the lowest number on the card, and 3.35 points per scoring opportunity. They sustain drives and almost never break one.

That combination usually produces exactly what our model projects: a controlled, low-scoring game rather than a blowout.

We project Michigan 27.5, Western Michigan 11.0 — 16.5 points on a total of 38.5. The books are up near twenty-six and a half, with SP+ at 25.5 and FPI 22.4. We are ten points below the market and the low number on the game.

THE READ

Two forces pull against each other here and our engine weights them differently to everyone else. It penalises Michigan for a first-year head coach, and it rewards Western Michigan for total continuity and a genuinely good defence. The market prices the talent gap and largely ignores both.

Worth noting the tension inside our own numbers: our board has Michigan 7th while AP have them 16th — a nine-place disagreement in Michigan's favour — and yet we project the narrowest margin of anyone in this game.

Toledo's defence allowed negative expected points per play last season — −0.078, on a 33.1% defensive success rate and 20.2% havoc. Those are not good MAC numbers; they are good numbers, full stop, better than any defensive unit Michigan State will face early. Our model has noticed, and it has Toledo winning outright.

BOTH PROGRAMMES

Both sides changed head coach, which is the complication running through this whole game.

Pat Fitzgerald takes over at Michigan State, returning to the sideline after seventeen years at Northwestern. He inherits a 4-8 team that did not win a road game, returning 21% of production, having lost 46 players to the portal against 28 arrivals — including four-star receiver Nick Marsh. Aidan Chiles is listed both as a returning starter under the new staff and among the outgoing transfers in our portal data, so his status is genuinely unsettled; his 2025 was strong, 99 total PPA at 0.431 a play.

Mike Jacobs takes over at Toledo from Mercer, where he went 20-6 with two playoff appearances at FCS level. He inherits far more: 66% of production returning, 98% of the passing, and Tucker Gleason back after 98 total PPA at 0.323. Toledo went 8-4 and were 6-0 at home.

THE NUMBERS

Toledo graded better on both sides of the ball. Offence: 0.204 expected points per play to Michigan State's 0.131, with a 19.3% stuff rate against 24.8% and 3.05 line yards against 2.58. Defence: −0.078 allowed against 0.198.

Our board still rates Michigan State marginally higher on raw rating — 59th to 69th — while projecting Toledo to 8.6 wins against Michigan State's 4.4. That is our model saying Toledo's profile is real but their schedule is soft.

We project Toledo to win by 6.9, 26.8 to 19.9, on a total of 46.7. The books have Michigan State by ten and a half. That is a seventeen-point disagreement, the largest on the entire Week 1 card.

THE READ

This is the game where our engine is furthest from the market anywhere this weekend, and the mechanism is clear: it is weighting 2025 efficiency heavily and the Big Ten name not at all. The market is doing close to the reverse. One of those is going to look foolish, and our own measured record says a seventeen-point disagreement is more often our error than the price's.

Mississippi State open at home against a UL Monroe side arriving from the Sun Belt, in a game where every available number lands within a few points of every other — and where both teams have to replace the man who threw the ball.

BOTH PROGRAMMES

Jeff Lebby is in year three in Starkville off 5-7. Blake Shapen generated 70 total PPA and Kamario Taylor 41 at a higher average of 0.275; Taylor is recorded as the returning starter, though Mississippi State's returning passing production reads essentially zero. What they do have is the 23rd-ranked recruiting class, the best of any team in this fixture by a wide margin, and 87% of their rushing production back.

Our board has them 42nd of 136 with 5.1 projected wins — the SEC schedule that follows this game is why that number is low, rather than any judgement about this particular afternoon.

Bryant Vincent is in year two at UL Monroe off 3-9, 1-6 on the road. They return 39% of production and, unusually, 78% of the passing — more continuity at the position than their hosts. Their problem is up front: a 24.5% stuff rate, among the highest on this weekend's board, on 2.74 line yards.

THE NUMBERS

UL Monroe's defence allowed 0.187 expected points per play and gives up explosive plays at 1.35. Their own offence produced 1.37 explosiveness — a boom-or-bust profile on both sides of the ball.

We project Mississippi State 42.1, UL Monroe 17.4 — 24.7 points on a total of 59.5. The books are near twenty-eight and a half. That is a gap of under four points, one of the closer agreements on the card.

THE READ

Little to separate the opinions here, and games like this are worth publishing precisely because they show what agreement looks like beside the double-digit disagreements elsewhere on this page. The recruiting gap is the honest argument for the market's slightly larger number.

Ohio went 8-4 and did not lose at home. They arrive in Lincoln against a Nebraska side whose quarterback situation our own sources cannot agree on, and the market makes it a three-and-a-half-touchdown game. We have it at ten and a half.

BOTH PROGRAMMES

Matt Rhule is in year four at Nebraska off 7-5. Dylan Raiola is recorded as a returning starter after 75 total PPA at 0.277 — and also appears on Nebraska's outgoing transfer list, and again on Oregon's incoming list. Those cannot all be true, so we are not naming a starter. Nebraska's returning passing production reads 5%, which is consistent with him having left.

Anthony Colandrea arrived through the portal along with four-star interior lineman Brendan Black. Nebraska return 46% of production overall and 74% of the receiving.

Ohio have a first-year head coach in John Hauser and return 62% of production — 94% of the passing, 100% of the rushing. Parker Navarro produced 117 total PPA at 0.290, though he also appears among the departures. Their offensive line is the standout number: a 13.0% stuff rate, the second-best figure of any team playing this weekend, on 3.32 line yards.

THE NUMBERS

Ohio's offence graded 0.207 expected points per play against Nebraska's returning uncertainty, and their defence allowed 0.125 with 14.9% havoc. This is a good MAC team, not a token opponent.

We project Nebraska 28.8, Ohio 18.2 — 10.6 points on a total of 47.0. The books are up near twenty-four. That is a thirteen-point disagreement on a game whose favourite is not in doubt.

THE READ

Our model is once again the low number on a Power Four side hosting a Group of Five opponent, and that pattern repeats through this entire week — six of our largest gaps share the shape. A pattern that consistent is more likely a property of our engine than six separate insights. What makes this one more defensible than most is the quarterback fog: a team whose passing production returns at 5% and whose starter may be in another conference is genuinely hard to price.

Ohio State are the unanimous choice of nobody and the near-unanimous choice of almost everyone: forty first-place votes, top of the AP poll, top of our board at +25.9. Ball State are 125th. The market asks for fifty points, and for once the interesting question is not who wins but what a game like this is actually worth reading about.

BOTH PROGRAMMES

Ryan Day is in year eight. Julian Sayin returns after leading the nation in completion percentage at 77.0% and posting 202 total PPA at 0.482 a play — the highest total of any quarterback playing this weekend. Ohio State return 68% of production and 94% of the passing, signed the fourth-ranked recruiting class, and their 2025 defence allowed negative expected points per play, the only such figure on this weekend's board.

Thirty-seven players left through the portal. It has changed nothing.

Ball State went 4-8 and, more tellingly, 0-7 on the road. Mike Uremovich is in year two. They return 6% of production; their rushing figure is deeply negative, which happens when the players who left graded below replacement — a statistical quirk, but one that tells you the run game was actively costing them points. Their offence produced −0.001 expected points per play and 2.71 points per scoring opportunity, both the worst numbers on the card.

THE NUMBERS

We project Ohio State 46.8 and Ball State to a negative figure, which gives a margin of 53.4. The books sit around fifty, SP+ goes to 60.3, FPI to 48.5.

That negative visitor projection is worth being explicit about: a football team cannot score below zero. It is our engine extrapolating past the range its inputs support, and we publish it flagged rather than quietly clipping it to a prettier number.

THE READ

All four numbers agree inside a twelve-point band around fifty, which is about as close to consensus as a game this lopsided gets. The genuine content here is not the margin — it is Sayin, who is the best returning quarterback in the sport by our data and worth watching for his own sake rather than the scoreboard's.

Oklahoma's 2025 defence allowed negative expected points per play — −0.016 — on a 30.7% defensive success rate with 22.7% havoc. Only Indiana graded better among teams playing this weekend. UTEP's offence produced 0.013 expected points per play on a 32.8% success rate, and their overall returning production is a negative figure.

BOTH PROGRAMMES

Brent Venables is in year five, John Mateer returns after 107 total PPA, and AP have Oklahoma 10th — exactly where our own board places them, one of the cleanest agreements between us and the voters anywhere in the top 25. They went 10-2 and were unbeaten on the road at 4-0.

Scotty Walden is in year three at UTEP off 2-10, winless away from El Paso. Malachi Nelson's 2025 produced 7 total PPA across the season and he has since left through the portal; EJ Colson arrives to replace him. UTEP's one distinguishing number is explosiveness at 1.45 — high, and the signature of a team that occasionally hit a long play and otherwise did very little, on 2.16 line yards and a 22.7% stuff rate.

THE NUMBERS

We project Oklahoma 41.2, UTEP 5.4 — 35.8 points on a total of 46.6. The books are near forty-one and a half, SP+ 40.6, FPI 36.9. We are the low number by roughly five points and everyone else is clustered tightly.

THE READ

Three of the four ratings land within a point of forty-one. Ours is the outlier and, unusually for this week, it is the conservative one. The mechanism is our engine crediting UTEP's explosiveness — a number generated by a handful of long plays against Conference USA defences, and the least likely thing to survive contact with a unit generating 22.7% havoc.

Oregon return 75% of last season's production, 94% of the passing, and the quarterback who generated it. In a sport that now rebuilds itself annually through the portal, that is close to an anomaly — and it is the single best reason to think an 11-1 team stays an 11-1 team.

BOTH PROGRAMMES

Dan Lanning is in year five off 11-1, unbeaten on the road at 5-0. Dante Moore returns having thrown for 3,565 yards at 71.8% with a 163.7 QBR, and his 181 total PPA at 0.382 a play is the highest of any quarterback in this weekend's fixtures. Oregon signed the third-ranked recruiting class in the country and added four-star safety Koi Perich and quarterback Dylan Raiola through the portal. Thirty-one players left; it did not matter.

Boise State are not a token opponent. They went 9-4, return 66% of production — 94% of the passing, 100% of the rushing — and Maddux Madsen is back after 88 total PPA at 0.280. Spencer Danielson is in year three. This is the Group of Five programme that most closely resembles a Power Four one.

THE NUMBERS

Oregon's offence graded 0.303 expected points per play on a 47.4% success rate, with 4.10 points per scoring opportunity. Their defence allowed 0.041 with 18.2% havoc. There is no weak side.

Boise State's own profile is respectable — 0.175 on offence, 0.100 allowed, a 35.3% defensive success rate — but their defensive explosiveness allowed is 1.43, the highest number on this card. Against an Oregon offence built on 1.34 explosiveness, that is the specific place this game is likely to break.

We project Oregon 38.6, Boise State 14.9 — a 23.7-point margin on a total of 53.5. The books sit within a point and a half of us on the spread. This is the closest our model and the market come to agreeing all weekend.

THE READ

On the game with the two clearest profiles — an elite, stable favourite and a good, known underdog — everybody lands in the same place. That is worth stating plainly, because most of this week's card has us disagreeing with the market by double digits. Where both sides of the roster are legible, our engine and the price converge.

AP have Oregon 2nd with fourteen first-place votes. Our board has them 2nd. No disagreement anywhere.

Matt Campbell spent ten seasons at Iowa State and won Big 12 Coach of the Year three times. He arrives at Penn State inheriting a 6-6 team that returns eight percent of its production, and he has brought his quarterback with him.

BOTH PROGRAMMES

Rocco Becht started three years for Campbell at Iowa State — 26 wins, 9,274 yards, 64 touchdowns — and follows him to Beaver Stadium. That continuity between coach and quarterback is rare for a first-year staff and is the single reason to think this transition goes smoothly.

Everything else is upheaval. Penn State's portal cycle was 38 in and 48 out, one of the heaviest net losses in the country, including five-star edge rusher Chaz Coleman and four-star tight end Luke Reynolds. Drew Allar and Ethan Grunkemeyer between them produced 99 total PPA in 2025 and both are gone; the returning passing production is 3%.

Marshall are the opposite in every respect. Tony Gibson is in year two, and they return 99% of production — 100% of the passing, 91% of the receiving, and a rushing figure above 100%. Carlos Del Rio-Wilson is back after 113 total PPA at 0.278. They went 5-7, but they are the same team that went 5-7, which is more than Penn State can say.

THE NUMBERS

Marshall's defence allows explosive plays at 1.40, one of the higher figures on the board, and that is the specific vulnerability a Campbell offence is built to attack. Penn State's own defence graded 0.105 expected points allowed per play with 15.9% havoc — respectable, not frightening.

We project Penn State 44.7, Marshall 20.7 — 24.0 points on a total of 65.4. The books are at twenty-four and a half. That is the closest our number and the market's come on any game this week; SP+ says 26.9 and FPI 25.0, so all four cluster inside three points.

THE READ

Total agreement on the margin, and an eleven-point gap on the total, where the market sits around fifty-four and a half. When our spread lands on the number, the disagreement has usually migrated to the total — and here our engine expects a Campbell offence to score more against a defence that surrenders big plays than the market does.

AP have Penn State 18th. Our board has them 34th — a sixteen-place gap, one of our largest disagreements with the poll anywhere.

Two of the longest-tenured head coaches in the sport meet in Pittsburgh: Pat Narduzzi in year twelve, Chuck Martin in year thirteen at Miami of Ohio. Between them that is a quarter of a century in the same two jobs, which in the portal era is close to archaeology.

BOTH PROGRAMMES

Pittsburgh went 8-4 and were better on the road (4-1) than at home (4-3). They return 40% of production and 59% of the passing, and their 2025 quarterback room was split between Mason Heintschel's 73 total PPA and Eli Holstein's 56 at a higher average of 0.369. Twenty-one players left, including four-star defensive lineman Francis Brewu — who has gone to Notre Dame — and linebacker Rasheem Biles.

Miami of Ohio went 7-6 and return 14% of production, with a passing figure marginally below zero. Dequan Finn's 66 total PPA has gone. Their numbers are balanced and unremarkable in both directions, which for a MAC side facing an ACC opponent is not the worst starting point: 0.102 expected points per play on offence, 0.088 allowed, 17.8% havoc — a higher disruption rate than Pittsburgh's own.

THE NUMBERS

Pittsburgh's edge is explosiveness at 1.34 against a Miami defence allowing 1.19 — they hit long plays and Miami do not surrender many. Miami's offensive line is the surprise of the fixture at 3.14 line yards, better than Pittsburgh's 2.83.

We project Pittsburgh 30.6, Miami (OH) 15.1 — 15.5 points on a total of 45.7. The books are at around sixteen, SP+ 12.7, FPI 16.1. Every number on this game sits inside a four-point band.

THE READ

The tightest cluster of opinions on the weekend. Our spread is within a point of the market and within a point of FPI, and the only gap is the total, where we are roughly five points under. When two settled programmes with no coaching change meet, our engine and the market stop arguing — which is itself a useful thing to notice about where our disagreements come from.

South Carolina went 4-8 and did not win a road game. Our board has them 18th of 136 at +11.2, which is the model stating plainly that the record understates the roster. AP do not rank them at all.

BOTH PROGRAMMES

Shane Beamer is in year six with LaNorris Sellers returning after 81 total PPA, and South Carolina return 69% of production across the board — 65% passing, 80% rushing, 67% receiving. That is genuine continuity, backed by the 14th-ranked recruiting class and a near-balanced portal cycle at 25 in and 24 out, though four-star tackle Josiah Thompson and receiver Vandrevius Jacobs both left.

Kent State are 130th on our board, the second-weakest side in this fixture set, and have promoted Mark Carney from offensive coordinator. But they return 64% of production and 103% of the passing, with Dru Deshields back after 64 total PPA — more continuity than most Power Four teams playing this weekend can claim.

THE NUMBERS

The gap is in the trenches. Kent State's offence gave up a 23.2% stuff rate on 2.43 line yards, both among the worst figures on this weekend's board — a line that loses the point of attack routinely. South Carolina's defence allowed 0.079 expected points per play with 16.0% havoc.

South Carolina's own offence is the curiosity: 0.105 expected points per play on a 38.2% success rate, with 1.32 explosiveness. That is a team that struggled to sustain drives and scored in chunks, which is the profile of a side that loses close games rather than one that is outclassed — and it explains a 4-8 record beside a top-20 rating.

We project South Carolina 34.3, Kent State 9.5 — 24.8 points on a total of 43.8. The books are up near thirty-six and a half, SP+ 37.5, FPI 32.1. We are the low number by more than ten points.

THE READ

Another Power Four host where our engine sits well below the field, and the pattern across this week is consistent enough to be a property of the model rather than six separate insights: it discounts blowouts, and against Group of Five opposition the market does not. Kent State's returning production is the one honest argument for our side of it.

South Florida went 9-3 and 6-0 in Tampa. Byrum Brown produced 178 total PPA at 0.409 a play, one of the most efficient quarterback seasons of any team on this weekend's board — and he is now at Auburn, having followed Alex Golesh there. Brian Hartline arrives from Ohio State to replace both of them.

BOTH PROGRAMMES

Hartline was Ohio State's offensive coordinator and wide receivers coach on the best offence in the country. It is his first programme, and the inheritance is thin: South Florida return 13% of production, 6% of the passing, and a rushing figure marginally below zero. Forty-four players arrived against thirty-four departures, including quarterback Luke Kromenhoek.

The 2025 offensive numbers are excellent and belong to people who have left: 0.298 expected points per play, a 46.7% success rate, 1.30 explosiveness.

Florida International were 7-5 under Willie Simmons in year two — a respectable record from a team our ratings put 116th. They return 32% of production with a passing figure marginally below zero of their own, and JJ Kohl arrived at quarterback. Their one strong number is 4.12 points per scoring opportunity, better than several Power Four teams playing this weekend.

THE NUMBERS

Florida International's offensive line is the weakness: a 22.5% stuff rate on 2.77 line yards. Both defences are mid-table — 0.150 allowed for FIU, 0.080 for South Florida.

We project South Florida 48.3, Florida International 23.7 — 24.6 points on a total of 72.0, the highest total we project anywhere this week. The books are at around twelve and a half, SP+ 15.3, FPI 14.1.

THE READ

A twelve-point gap on the side and a seventeen-point gap on the total. Our engine is projecting a shootout nobody else sees, on the back of an offensive rating produced by a coordinator who has left and a quarterback who is now in the SEC. This is the same structural error as our Cincinnati number, in a more extreme form.

Miami arrive in Palo Alto ranked seventh in the country. Stanford went 4-8, lost all six road games, return 21% of their production and have handed the programme to a coach who has never run one.

BOTH PROGRAMMES

Tavita Pritchard comes from the Washington Commanders, where he coached quarterbacks. It is his first head coaching job, and the inheritance is thin: Stanford's returning passing production is a negative figure, their portal cycle was the quietest of any team playing this weekend at six in and twelve out, and their offence graded 3.02 points per scoring opportunity — the lowest number on the card.

They did sign the 37th-ranked recruiting class, which is the one genuinely encouraging line in the profile.

Mario Cristobal is in year five at Miami off 10-2, 7-1 at home. They lost Carson Beck's 152 total PPA and replaced him with Darian Mensah, a four-star portal arrival our data describes as instantly among the better signal-callers in the sport. Miami return only 6% of their passing production but 91% of the rushing, and signed the ninth-ranked class.

THE NUMBERS

Miami's defence is the story: 0.034 expected points allowed per play with 18.9% havoc, and 4.36 points per scoring opportunity at the other end. Stanford's defence allowed 0.178 with 14.6% havoc, and their offensive line gave up a 18.6% stuff rate on 2.62 line yards.

There is no phase of this game where Stanford grade better.

We project Miami 35.1, Stanford 7.5 — 27.6 points to the visitors on a total of 42.6. The books are near twenty-three, SP+ 21.8, FPI 22.7. All three outside numbers sit within a point of each other, and ours is five points above all of them.

THE READ

That convergence is the thing to weigh. When the market, SP+ and FPI agree to within a point, they are usually not all missing the same thing — and our number is the outlier that needs justifying. What our engine is doing is compounding Stanford's road record, their returning production and a first-year head coach into a projection of near-total collapse. Each input is defensible; multiplying all three together is where preseason models overreach.

AP have Miami 7th. Our board has them 12th.

Texas State's offence produced 0.311 expected points per play last season. Texas produced 0.179. That is not a typo, and it is the reason our number on this game sits fifteen points below the market's.

BOTH PROGRAMMES

Steve Sarkisian is in year six with Arch Manning returning — 3,163 yards, 26 touchdowns, seven interceptions and ten rushing scores in 2025, 105 total PPA at 0.253 a play, and a Heisman favourite carrying a reported $5.4m NIL valuation. Texas return 72% of production, 85% of the passing, and were 6-0 at home. They brought in five-star receiver Cam Coleman and lost 32 players to the portal against 22 in, and signed the seventh-ranked recruiting class in the country.

Texas State went 6-6 under G.J. Kinne, in year four, and return 76% of production — more than Texas — including 100% of their passing. Brad Jackson is back after 172 total PPA at 0.334 a play, a higher total than Manning's. Their offence graded 4.42 points per scoring opportunity against Texas' 3.52.

None of which makes them favourites. It does make them a genuinely awkward opener.

THE NUMBERS

Texas' strength is the defence: 0.098 expected points allowed per play, a 39.2% defensive success rate and 17.9% havoc. Texas State's defence is the weak unit here at 0.158 allowed.

The Texas offensive profile is the curiosity — a 22.2% stuff rate and 2.83 line yards, both mediocre, on 1.37 explosiveness. That is a team that struggled to run and scored in chunks.

We project Texas 42.3, Texas State 26.8 — 15.5 points on a total of 69.1. The books have Texas by around thirty and a half, with a total in the low fifties. We are fifteen points below on the side and roughly eighteen points above on the total.

THE READ

Both halves of that disagreement come from the same source: our engine believes Texas State can move the ball. If that is right, this is a shootout Texas win by two or three scores. If it is wrong — if a Sun Belt-calibre offence that has moved to the Pac-12 cannot function against SEC defensive speed — then the market's thirty is closer and our total is badly high.

AP have Texas 5th. So does our board.

Texas A&M went 11-1 with a perfect home record, and AP have them 8th. Our board is higher at 6th. Missouri State arrive from Conference USA with a first-year head coach hired off SMU's offensive staff.

BOTH PROGRAMMES

Mike Elko is in year three with Marcel Reed returning after 141 total PPA at 0.312. The defence is the strength and it is a serious one: a 33.8% defensive success rate and 21.4% havoc, both among the best figures of any team playing this weekend.

Casey Woods takes over at Missouri State having been SMU's offensive coordinator, and inherits a 7-5 team that was better on the road (4-2) than at home (3-3). They return 31% of production and 5% of the passing — Jacob Clark's 103 total PPA is gone — with nineteen portal arrivals against eleven departures, one of the few net-positive cycles on the board.

THE NUMBERS

Missouri State's profile is the most extreme on the card in both directions. Their explosiveness is 1.49, the highest of any team playing this weekend — and their stuff rate is 27.3%, also the highest, on 2.47 line yards. That is an offence that gets stopped behind the line more often than anyone and, when it does not, goes a long way. Against a defence generating 21.4% havoc, the first half of that pairing is the one that matters.

We project Texas A&M 47.6, Missouri State 14.3 — 33.3 points on a total of 61.9. The books are near thirty-nine and a half, SP+ 42.5, FPI 32.8. We sit between FPI and the rest.

THE READ

For once our number is not the outlier — FPI is lower than us and SP+ is nine points higher. The total is where we separate, projecting nearly ten points more football than the market, which comes from crediting Missouri State's explosiveness without fully pricing how often that offence simply does not get started.

Sam Houston's 2025 record carries an oddity worth flagging before anything else: their verified home record is 0-0. They went 2-10 with every logged result on the road or at a neutral site, which is unusual enough that it should temper how much weight anyone puts on the profile.

BOTH PROGRAMMES

Phil Longo is in year two. Sam Houston are 135th of 136 on our board at −14.2, second-lowest of any team playing this weekend. Their offence produced 0.001 expected points per play — as close to nothing as the metric gets — on a 25.7% stuff rate and 2.79 points per scoring opportunity, and their defence allowed 0.237 on a 47.9% defensive success rate. Both are among the worst figures on the card.

They do return 58% of production, though the passing figure is negative — Hunter Watson's 11 total PPA across a full season tells you why.

Troy were 8-5 under Gerad Parker, in year three, balanced 4-2 at home and 4-3 away, with 7.0 projected wins on our board. Tucker Kilcrease returns after 40 total PPA, and Troy's returning passing and rushing production are both above 100%. Fourteen players arrived against twenty-seven departures.

THE NUMBERS

We project Troy 40.2, Sam Houston 16.6 — 23.6 points on a total of 56.8. The books are near sixteen, ESPN's number nearer seventeen and a half, SP+ 20.5, FPI 13.6. We are the highest number on the game by three points over SP+ and seven over the market.

THE READ

The four ratings span ten points, which is wide for a game between a Sun Belt side and a bottom-five opponent, and our engine is the most bullish of them. The missing home-game history is a genuine reason to hold our number lightly: a team whose entire season was played away from home has a profile that no model is reading correctly.

Oklahoma State's returning production for 2026 reads zero percent. Not two, not six — zero overall, zero passing, zero rushing, zero receiving. After a 1-11 season, Eric Morris has replaced the entire football team, and the market makes that rebuilt team a two-touchdown favourite on the road.

BOTH PROGRAMMES

The churn is the largest in the sport: 55 players in, 65 out. Morris arrives from North Texas and has brought his quarterback with him — Drew Mestemaker, a five-star portal move who threw for 4,379 yards and 34 touchdowns at North Texas in 2025 and generated 162 total PPA there. Four-star running back Caleb Hawkins and receiver Wyatt Young followed the same path.

That is the market's entire case, and it is a reasonable one: the 1-11 record belongs to a roster that no longer exists, and the quarterback belongs to a system he already knows.

Our engine sees the numbers that remain. Oklahoma State's 2025 offence produced −0.012 expected points per play on a 34.2% success rate; their defence allowed 0.261, the second-worst figure on this weekend's board.

Tulsa were 4-8 under Tre Lamb, in year two, and graded better in both directions. They return 41% of production, 69% of the passing.

THE NUMBERS

We project Tulsa 28.1, Oklahoma State 20.6 — Tulsa by 7.5 on a total of 48.7. The books have Oklahoma State favoured by around fourteen, SP+ 10.5, FPI 9.8. That is a twenty-point swing between our number and the market's, and we are alone on our side of it.

THE READ

This is the cleanest test on the card of a question every preseason model faces: what is a new coach and a transferred quarterback worth? Our engine prices it at close to nothing and leans entirely on 2025 efficiency. The market prices it at roughly three touchdowns of improvement.

Given Oklahoma State return zero percent — a figure our model is effectively ignoring while projecting last year's efficiency forward — the honest reading is that our number is the one built on the weaker assumption here.

USC signed the number one recruiting class in the country. They open tied fourteenth in the AP poll with BYU, both on exactly 839 points — one of only two joint rankings in the top 25. And they do it without a settled quarterback.

BOTH PROGRAMMES

Lincoln Riley is in year five off 9-3, with a home-road split as stark as any on the board: 7-0 at the Coliseum, 2-3 away from it. They open at home.

Jayden Maiava produced 205 total PPA at 0.477 a play in 2025, among the best seasons in this fixture set, and our data lists him as the presumed starter while flagging it unconfirmed — so we are not treating the job as settled. Husan Longstreet, who averaged 0.851 PPA in relief, has left through the portal. USC return 59% of production and 91% of the passing, and their portal cycle was quiet by modern standards: nine in, twenty-two out.

Fresno State are a solid, unspectacular 8-4 side under Matt Entz in year two, balanced 4-2 home and away. E.J. Warner's 126 total PPA at 0.424 has gone and their returning passing production is 7%; what comes back is the receiving corps at 70%.

THE NUMBERS

USC's offence graded 0.344 expected points per play on a 49.0% success rate with 1.34 explosiveness — top-tier across the board. The defence is the softer half at 0.149 allowed and a 44.8% defensive success rate, the weakest defensive success rate of any ranked team playing this weekend.

Fresno State's defence is genuinely decent: 0.083 expected points allowed per play with 17.4% havoc, better than USC's own numbers.

We project USC 34.8, Fresno State 16.3 — 18.5 points on a total of 51.1. The books are near twenty-three, SP+ 21.6, FPI 22.0, all clustered. We are roughly four points below the field, and our total is within half a point of the market's — rare on this card.

THE READ

A quiet game by this week's standards, and the AP tie is the more interesting fact about it. The thing to watch is USC's defence: a unit allowing a 44.8% success rate is the profile of a team that wins shootouts, and Fresno State are equipped to make it one for a half.

Rich Rodriguez is in the second year of a return to Morgantown that took seventeen seasons to arrange. West Virginia went 4-8 and return 14% of last season's production. Coastal Carolina went 6-6 and return 15%. Neither team is bringing much of itself into this game.

BOTH PROGRAMMES

West Virginia's returning passing production is the one healthy figure at 62%, with Nicco Marchiol and Khalil Wilkins both back — though between them they generated only 35 total PPA last season, which is a low base to build from. They did sign the 25th-ranked recruiting class, comfortably the best of any team in this fixture.

Ryan Beard arrives at Coastal Carolina from Missouri State. His portal cycle went 32 in and 43 out — a net loss of eleven — and the offence he inherits produced 0.062 expected points per play on a 36.1% success rate, one of the weaker pairings on the card. What Coastal do have is 3.80 points per scoring opportunity, which says they converted the chances they got.

THE NUMBERS

Neither defence is good. Coastal allowed 0.203 expected points per play with 15.3% havoc; West Virginia allowed 0.159 with 17.7%. West Virginia's offence graded 0.071 — barely above Coastal's — on 2.63 line yards.

There is very little in the underlying numbers to justify a three-touchdown line, which is precisely our engine's argument.

We project West Virginia 34.6, Coastal Carolina 25.7 — 8.9 points on a total of 60.3. The books are up near twenty-one and a half, SP+ 15.9, FPI 14.8. We are six points below FPI and nearly thirteen below the market.

THE READ

Same shape as several games this week: a Power Four host, a Group of Five visitor, and our engine projecting a far closer game than anyone else. What makes this one more interesting is that West Virginia's own 2025 numbers were poor — this is not a good team giving three touchdowns, it is a 4-8 team doing so, and our model is the only number on the board treating it that way.

California and UCLA open against each other with two brand-new head coaches, and the coaching carousel behind them is tangled enough to be worth a paragraph on its own.

BOTH PROGRAMMES

Bob Chesney takes over at UCLA from James Madison, where he went 21-6 across two seasons. James Madison replaced him with Billy Napier, who had been fired mid-season at Florida — and Florida in turn hired Jon Sumrall from Tulane. Four programmes, one chain, all of it playing out in this weekend's fixtures.

Tosh Lupoi arrives at California from Oregon, where he was defensive coordinator on one of the best defences in the country.

UCLA went 3-9 and return 57% of production. Nico Iamaleava's 88 total PPA at 0.227 sits behind a 55% returning passing figure, and their portal cycle was aggressive: 42 in against 29 out, including four-star linebacker Sammy Omosigho and receiver Aidan Mizell.

California were 7-5 and return 58%, with 89% of the passing — Jaron-Keawe Sagapolutele back after 102 total PPA. They lost 39 players against 32 arrivals, including four-star linebacker Cade Uluave.

THE NUMBERS

The gap our model sees is defensive. UCLA's 2025 defence allowed 0.261 expected points per play on a 49.8% defensive success rate with just 12.0% havoc — the weakest defensive profile of any team playing this weekend. California allowed 0.125 with a better success rate on both sides.

Neither offence is good. California managed 3.70 points per scoring opportunity, UCLA 2.98.

We project California 30.8, UCLA 21.9 — California by 8.9 on a total of 52.7. The books have it within a point and a half of pick'em, with UCLA marginally favoured at two of the three.

THE READ

An eight-point disagreement on a game the market calls a coin flip. Our engine is leaning almost entirely on UCLA's defensive numbers, which were genuinely poor, against a market that sees two 2026 rebuilds and declines to separate them.

Worth noting: our own board has UCLA rated higher than California — 48th to 60th — while the projection favours California by nearly nine. That is the home-field adjustment and the defensive matchup doing the work, and it is a case where our rating and our projection point in different directions.

Every rating available has this game close, and closer to each other than on any other fixture this week. What they are agreeing about is less obvious than it looks.

BOTH PROGRAMMES

Dan Mullen's UNLV went 10-3, and the engine of it has gone: Anthony Colandrea produced 183 total PPA at 0.353 a play — one of the best quarterback seasons on this weekend's board — and is now at Nebraska. UNLV return 15% of production and 4% of the passing.

Their 2025 offensive numbers are excellent and increasingly historical: 0.298 expected points per play, a 46.6% success rate, 1.36 explosiveness and 4.45 points per scoring opportunity, the second-best conversion figure of any team playing this weekend.

Hawaii are the more intact side. Timmy Chang is in year five off 8-4, built at home — 6-1 there against 2-3 on the road — and they return 64% of production with 91% of the passing. Micah Alejado is back after 133 total PPA. Sixteen arrived, nine left, the smallest outflow of any team on this card.

The travel involved in playing in Honolulu is a real and frequently discussed factor, and it falls on UNLV.

THE NUMBERS

UNLV's defence is the weaker unit at 0.210 expected points allowed per play; Hawaii's allowed 0.163. Neither generates much disruption.

We project UNLV by 1.5, 33.4 to 31.9, on a total of 65.3. The books have UNLV by around two and a half to three and a half, SP+ 3.7, FPI 1.7. All five numbers sit within two and a half points of each other.

THE READ

The tightest agreement on the entire Week 1 card. Where our model and three independent ratings converge this closely, the useful information is elsewhere: our total is 65.3 against a market near fifty-eight and a half, and the case for it rests on two offences that were genuinely productive last season — one of which no longer has the man who made it so.

Tyson Helton is in his eighth season at Western Kentucky, which in the current sport is close to a monument. They went 8-4 and return 84% of last season's production — 95% of the passing, 85% of the receiving — with Maverick McIvor back after 81 total PPA. Both the market and our model favour the visitors; they disagree sharply about by how much.

BOTH PROGRAMMES

Western Kentucky's defence is the stronger half of the profile: 0.096 expected points allowed per play on a 36.4% defensive success rate with 16.1% havoc. The offence converts well too, at 4.21 points per scoring opportunity.

Nevada were 3-9 under Jeff Choate, in year three, and are 120th on our board. Carter Jones generated 29 total PPA across the season — one of the lowest quarterback figures of any team playing this weekend — and their offence produced 0.034 expected points per play, among the weakest on the card.

THE NUMBERS

We project Western Kentucky by 12.2, 29.8 to 17.6, on a total of 47.4. The books have them by around two and a half, ESPN's number is nearer four and a half, SP+ 1.9, FPI 4.6. We are nine points clear of the market and more than ten clear of SP+.

THE READ

Three outside ratings put this inside a touchdown; we have it at nearly two. The gap comes almost entirely from how far our engine downgrades Nevada's offensive efficiency — which was genuinely poor, but poor against a Mountain West schedule, and our model does not adjust for that as sharply as the market does.

The continuity argument is on our side: an 84%-returning team against one that won three games is the sort of matchup where last season's numbers should carry more weight than usual.

New Mexico went 9-3 last season and did not lose at University Stadium — 6-0 at home, 3-3 on the road. Our board has them 109th of 136. That gap between record and rating is the most interesting thing about this game.

BOTH PROGRAMMES

Jason Eck is in year two, and the profile is solid rather than spectacular: 0.160 expected points per play on offence, 0.105 allowed, 17.2% havoc. They return 58% of production with 74% of the passing — Jack Layne back after 88 total PPA — and ran an even portal cycle at fourteen in and fourteen out.

Our rating of 109th with 6.5 projected wins is a schedule judgement rather than a quality one: a 9-3 record built against a Mountain West slate does not move our engine the way it moves a poll.

Central Michigan went 7-5 under Matt Drinkall, in year two, and were 4-1 at home. They return 54% of production, with a rushing figure above 100% but only 9% of the passing — Joey Labas' 70 total PPA has gone. Their defence is the better half at 0.084 expected points allowed per play with 16.9% havoc, a stronger disruption rate than New Mexico's.

THE NUMBERS

Neither offence is explosive: Central Michigan 1.16, New Mexico 1.25. Both convert respectably — 3.60 and comparable points per scoring opportunity — which is the profile of two teams that grind rather than strike.

We project New Mexico 25.9, Central Michigan 16.0 — 9.9 points on a total of 41.9. The books are near twelve and a half, FPI 11.8, SP+ 17.6. We are the low number, three points under the market.

THE READ

Broad agreement other than SP+, which sits roughly eight points above us and five above the market. The larger disagreement is the total: 41.9 against a market near forty-nine, driven by two offences that neither strike quickly nor return much of their passing production.

They are playing this one at Lambeau Field, which is the sort of detail that sells a television window on its own. The football underneath it is more lopsided than the venue suggests.

BOTH PROGRAMMES

Notre Dame went 10-2 and return 51% of production, including 91% of the passing. Marcus Freeman is in year five. Their offence produced 0.333 expected points per play with 1.50 explosiveness — the highest explosiveness figure of any team playing this weekend — and 4.79 points per scoring opportunity, also the best on the card. They signed the fifth-ranked recruiting class and lost only seven players to the portal.

Wisconsin went 4-8 and did not win a road game. Luke Fickell is in year four. Their offence managed 0.022 expected points per play and 2.28 points per scoring opportunity, the weakest pairing in this fixture set. Their returning passing production is a negative number — a quirk that occurs when the quarterbacks who left graded below replacement, and one that says more about last season than this one. The quarterback job is unsettled; our source flags it for verification, so we are not naming a starter.

Thirty-three players arrived through the portal and twenty-six left. That is a squad being turned over, not tuned.

THE NUMBERS

Notre Dame's stuff rate on offence is 22.9% — high, and the one soft spot in their profile. Wisconsin's defence is the unit that would have to exploit it, and it allowed 0.173 expected points per play on a 44.8% success rate with 12.9% havoc, the lowest havoc rate on the card. They do not create disruption.

We project Notre Dame 34.4, Wisconsin 7.2 — 27.2 points on a total of 41.6. The books are around twenty. We are seven points more emphatic than the market.

THE READ

Our engine is taking Wisconsin's offensive collapse at face value and projecting almost nothing. The market is pricing some regression toward the mean, which is the more common bet for a Big Ten programme with a fourth-year coach and thirty-three new players. Our number requires last season's worst-in-class offence to still be that, in September, after a roster overhaul.

AP have Notre Dame 4th with six first-place votes. So do we.

Ole Miss went 11-1 and 8-0 at home. Louisville went 8-4. Both arrive in Nashville having lost the person most responsible for that, and both have promoted or imported a replacement who has never done this job.

BOTH PROGRAMMES

Pete Golding takes over at Ole Miss, promoted from defensive coordinator after Lane Kiffin left for LSU. It is an internal hire, which our engine treats more gently than an outside one, and he keeps Trinidad Chambliss — whose 194 total PPA at 0.345 a play was among the best quarterback seasons in the country. Ole Miss return 50% of production and 79% of the passing.

That continuity is real, but the edges have been stripped: five-star edge rusher Princewill Umanmielen left for LSU, quarterback Austin Simmons departed after a 0.526 average PPA in relief, and 25 players went out against 28 in.

Louisville are the more thoroughly rebuilt side. Jeff Brohm is in year four, but the passing game returns zero percent of last season's production — Miller Moss is gone, and Lincoln Kienholz arrives from Ohio State to replace him. Thirty-nine players left, thirty-three arrived. What returns is the run game, at 85%.

THE NUMBERS

Ole Miss' offence was excellent: 0.307 expected points per play, a 48.5% success rate and 4.44 points per scoring opportunity. Louisville's defence is the reason this is not a formality — 0.060 expected points allowed per play on a 36.4% defensive success rate, better than Ole Miss' own defensive numbers.

We project Ole Miss 33.7, Louisville 23.4 — 10.3 points on a total of 57.1. The books have it at six and a half. We are roughly four points clear of the price on the favourite.

THE READ

Our model docks Louisville for the coaching continuity it does have and for a passing attack starting from zero, and it likes Ole Miss' returning quarterback more than the market does. The counter-argument is on the field: Louisville's defence graded better than Ole Miss', and defences travel more reliably than first-year passing games.

AP have Ole Miss 9th and Louisville 24th. Our board has them 11th and 37th — the Louisville gap is thirteen places, one of our largest disagreements with the poll this week.

The Apple Cup opens both seasons, and the roster arithmetic runs against the ranking. Washington State return 91% of last season's production — 100% of the passing, 100% of the rushing — under a first-year head coach. Washington, ranked 17th, return 54%.

BOTH PROGRAMMES

Kirby Moore arrives at Washington State from Missouri, where he was offensive coordinator. He inherits a 6-6 team that was 5-1 at home and 1-5 away, and, unusually for a new coach, an almost entirely intact roster: Zevi Eckhaus and Jaxon Potter both return, and 28 players arrived against 41 departures without touching the core.

Jedd Fisch is in year three at Washington off 8-4. Their quarterback position is the muddle: Demond Williams Jr. produced 142 total PPA at 0.332 and appears in our data both as the presumed returning starter and on the outgoing transfer list. Those cannot both be true, and until one is resolved we are not naming a starter. Washington signed the 13th-ranked recruiting class.

THE NUMBERS

Washington's offence is clearly the better unit: 0.290 expected points per play on a 48.9% success rate, against Washington State's 0.126 and 41.6%. Their defence allowed 0.057 with a 38.0% success rate.

Washington State's counter is durability rather than quality — 3.93 points per scoring opportunity, respectable, and a defence allowing just 1.08 explosiveness, the lowest figure of any team playing this weekend. They do not give up long plays.

We project Washington 29.8, Washington State 12.2 — 17.6 points on a total of 42.0. The books are near twenty-two and a half, SP+ 22.6, FPI 16.4. We sit between FPI and the rest, which is unusual for us this week.

THE READ

A rivalry game with a new coach on one side is the situation preseason models handle least well, and for once our number is inside the range of the outside ratings rather than beyond it. The tension worth watching: Washington's offence is built on explosive plays, and Washington State's defence is the best on this card at preventing them.

AP have Washington 17th; our board has them 16th. Close agreement, for once.

SMU are favoured on the road in Tallahassee, which is one of the few away-favourite lines of the weekend, and Florida State's summer explains why.

BOTH PROGRAMMES

Mike Norvell is in year seven off 5-7, and the split is stark: 5-2 at Doak Campbell, 0-5 on the road. The roster has been gutted and reassembled. Forty-three players left through the portal against twenty-two arrivals — one of the heaviest net losses in the country — including four-star edge rushers Mandrell Desir and Darryll Desir and corner Ja'Bril Rawls. Running back Quintrevion Wisner arrived from Texas and offensive tackle Xavier Chaplin from elsewhere, both four-star.

Florida State return 40% of production but only 12% of the passing: Thomas Castellanos and his 133 total PPA are gone, and the succession is unsettled — our source flags competing reporting on who takes over, so we are not naming a starter. They did sign the 16th-ranked recruiting class, the highest of any team playing this weekend.

Rhett Lashlee is in year five at SMU off 8-4, returning 57% of production and 97% of the passing. Kevin Jennings is the presumed starter — unconfirmed in our data — after 151 total PPA at 0.331.

THE NUMBERS

This is closer than the reputations suggest. Florida State's offence actually graded higher than SMU's — 0.301 expected points per play against 0.249, with a 12.9% stuff rate that is the best figure on this weekend's board and 3.47 line yards to match. Florida State can run the ball.

SMU's edge is defensive: 0.033 expected points allowed per play against Florida State's 0.148, with 18.8% havoc. That is the difference in the game and the reason SMU are priced up.

We project SMU 27.7, Florida State 23.6 — SMU by 4.1 on a total of 51.3. The books have SMU by two and a half. We are on the same side and slightly firmer.

THE READ

Both our model and the market see the same thing: a better SMU team, a real but not decisive Florida State home advantage, and a game inside a touchdown. Where our engine adds something is the run-game mismatch — Florida State's line was among the best in the country at avoiding stuffs, and SMU's defence is the unit built to test that.

AP have SMU 19th and Florida State unranked. Our board has them 20th and 29th, which is the closest agreement with the poll we have on any ACC team this week.