Oklahoma at Michigan. Lean: Michigan +5.5. Lean: Under 43.5.

Spread.Michigan +5.5 is a modest price-based lean. Our calibrated Michigan +3.7 provides only 1.8 points of apparent value, insufficient to establish an edge given the model’s record. More persuasive is the seven-point move from Michigan -1.5 in the captured preseason market: Oklahoma’s superiority and Michigan’s offensive problems already command a substantial premium. Michigan’s defense allowed Western Michigan only 221 yards despite an ugly offensive performance, offering a credible route to staying within one score. Oklahoma’s protection remains a variable: Maikkula and Fodje reportedly practiced fully, but Fasusi’s shoulder status remains unresolved. Their expected returns weaken the home-side argument; they should not be treated as confirmed absences. Underwood’s struggles and Oklahoma’s explosive passing are substantial reasons to favor the visitors outright. Nevertheless, one opener is insufficient to assume Michigan’s offensive floor is permanent. I take the points with limited conviction, without interpreting reported public support for Oklahoma as proof of sharp Michigan money.

Total.Under 43.5 is the stronger, though still moderate, lean. Michigan’s offensive uncertainty and credible defensive showing support a game in which sustained scoring drives are difficult. Oklahoma’s unresolved tackle situation adds downside to its offensive efficiency. The over case deserves weight: Mateer threw three touchdown passes in the opener, returning interior linemen could improve protection, and Michigan turnovers or Sategna’s return ability can create points without long drives. Weather looks playable, so this is not a weather under. Moreover, the captured total has already fallen from 45.5-46.5 at FanDuel and DraftKings, reducing value. Our 41.43 projection offers only a small cushion and receives little independent weight. I favor defensive resistance over explosive scoring, while acknowledging that verified tempo data and final availability remain unavailable.

Where our number looks wrong.Michigan’s prior may overstate its present offensive competence under a new staff; the final 13-12 score also hides dependence on a last-play touchdown. Conversely, treating Oklahoma’s 51 points against UTEP as repeatable offensive strength would overreact to opponent quality and a punt-return touchdown. The totals documentation contradicts itself: it says no weather input while explicitly subtracting 1.97 points for wind. A forecast near 9-10 mph does not independently justify that much suppression. Neither model captures the current offensive-line uncertainty.

Spread.Our number is Michigan +3.7 against a 5.5 market, a 1.7-point gap our own model cannot cash (46-50% ATS in every bucket), so the case has to come from the news. The market moved 7 to 9 points off two low-information games: DraftKings and FanDuel opened Michigan -1.5 (our captures 9/1 and 8/30), the number peaked at Oklahoma -6.5 Monday and sits -5.5. It got there because Michigan needed a Hail Mary to beat Western Michigan 13-12 while Oklahoma hung 51-0 on UTEP. Both readings are noisy. WMU was the media pick to win the MAC and never reached the end zone; Michigan's defense allowed 221 yards and four field goals, and the offense lost that game with three turnovers and nine penalties, the variance items. Oklahoma travels with left tackle Michael Fasusi uncertain (SI, 9/8, which reports OU may save him for Georgia), its center and right guard back only this week from camp knee injuries, and RB Tory Blaylock likely out. 81% of Action Network tickets are on Oklahoma (9/9) while the number ticked back from -6.5. Take the points in a 43-point game.

Total.Weather removes the excuse either way: mostly sunny, high 82, no meaningful wind at a noon kick. The under is a defense argument. Oklahoma held UTEP to 198 yards and 2-of-13 on third down; Michigan held a MAC favorite to 221 yards and no touchdowns. Michigan managed 13 points on 276 yards with a 30% third-down rate, and Oklahoma's line arrives without its starting left tackle and probably without Tory Blaylock. Last year's meeting finished 24-13, 37 points. The honest deduction: the market has already made this move.

Where our number looks wrong.Our rating is still mostly the preseason prior, so it credits Michigan with a 15.2 built on recruiting composite and returning production while knowing nothing about the actual problem: the quarterback and a new offensive scheme. It cannot see that Underwood went 12-of-22 with a 52.9 QBR and has a head coach publicly refusing to guarantee his job, and it equally cannot see that Oklahoma's front travels without its starting left tackle and with two other starters a week off knee injuries. Flat 2.5-point home field also undersells a noon kickoff in a 110,000-seat stadium against an SEC team on an 11 a.m. body clock. On balance the 3.7 is probably too kind to Michigan on merit; I am taking the side anyway because a 7-to-9-point market move built on UTEP and one MAC game is a bigger error than our model's, not because our number is right.

Spread.I'm overriding the model's pass. Week 1 gave real, not cosmetic, information: Michigan needed a missed WMU field goal to escape 13-12, with QB Bryce Underwood shaky (12/22 passing) and the run game stuck at 3.69 YPC, players and beat writers called it 'soft,' not just an ugly final score. Oklahoma pitched a 51-0 shutout with John Mateer, fully healed from last year's hand injury, efficient through the air (225 yds, 3 TD). Oklahoma's OL news (Fasusi shoulder, Maikkula back from an MCL sprain) is the main complication, but Venables sounds optimistic, and Oklahoma's defensive front looks like the bigger factor against a Michigan offense that hasn't shown it can move the ball against a real defense. Take Oklahoma -5.5.

Total.The market itself has independently drifted down about 2-3 points from its late-August/early-September opens (44-46.5, per line history) to 43-43.5 now. Weather is a non-factor: mild and dry, high near 81-82°F, wind ~9mph, under 5% rain chance. Both offensive lines have real questions, Michigan's offense looked inefficient in Week 1 (13 points, 3.69 YPC) and Oklahoma's OL (Fasusi, Maikkula) is unsettled, which points toward stalled drives and field position more than a shootout. Lean Under 43.5.

Where our number looks wrong.The in-season rating is one game deep and still mostly the preseason prior (recruiting composite, returning production, portal net), so it doesn't fully credit what Week 1 tape actually showed: a Michigan offense that looked genuinely limited (not just an artifact of a 1-point final) against a MAC defense, versus an Oklahoma team that looked SEC-caliber on both sides. The model also has no mechanism for trench-level injury detail (Fasusi/Maikkula for Oklahoma, Babalola for Michigan), exactly the kind of matchup-specific news beat coverage is flagging this week that a team-rating number can't see.