Colorado at Northwestern. Lean: Colorado +3.5. Lean: Under 48.5.

Spread.I prefer Colorado +3.5 because the Buffaloes have already demonstrated meaningful improvement against an FBS opponent: they won at Georgia Tech while holding its rushing attack to 51 yards on 30 attempts. Julian Lewis struggled for much of that game but subsequently produced 366 yards and four touchdowns against Weber State, confirming upside without proving consistency. Northwestern’s 497-yard opener was explosive but flattered by two long plays and a late pull-away after the game stood 20-18. More importantly, starting center Jackson Carsello suffered a neck injury, and David Braun declined to disclose his Colorado-game status. The market has already recognized Colorado’s improvement and that uncertainty, moving Northwestern from as high as -5.5 to -3.5; reported Colorado-heavy handle therefore is not an unpriced signal. Still, getting the key 3.5 with the more tested defense is preferable in a matchup I price close to Northwestern -2.5.

Total.The model’s 47.8 is effectively identical to 48.5, so this is a modest price-based under. The total has risen roughly three points from widely displayed 45.5 openers after both teams posted explosive FCS production, while 76% of reported tickets and 83% of money were on the over. Against its lone FBS opponent, Colorado played a 14-13 game, ran 49 times and possessed the ball for 34:09. Northwestern’s opener was only 20-18 late despite 497 yards, showing how a few explosive plays inflated its final score. The countercase is legitimate: Lewis, Chiles and Chip Kelly create big-play upside, while Northwestern’s red-zone stops could regress. Possible showers are only a slight under input because winds should remain manageable. Carsello’s uncertain status breaks the tie toward under.

Where our number looks wrong.The model’s -4.1 treats a 0.6-point difference as meaningful when its ATS testing says it is not. It cannot properly value Colorado’s rebuilt run defense, new offensive structure, Northwestern’s bye, or Carsello’s unresolved status. It also risks overweighting Northwestern’s 16-point final margin in a game that was 20-18 late. My adjusted number is Northwestern -2.5.

Spread.Our number (home -4.1) and the market (-3.5) essentially agree, so this is a coin-flip spot our model flags as a pass. What tips me to Colorado is the roster and the line move. Colorado is 2-0 with a quality road win at Georgia Tech (14-13, a Power 4 win in a low-scoring grind) and has allowed just 17 ppg; their four-year blue-chip ratio (0.31) dwarfs Northwestern's (0.05). Northwestern's lone data point is a 34-18 home win over FCS South Dakota State. That reverse move plus Colorado's defense and proven road toughness make +3.5 the side. The catch is Colorado's returning production is only 14% and QB Julian Lewis is a redshirt freshman, so the sample is thin and volatile. I take the points, at a lean, not a strong bet.

Total.The total has been bid up hard, from a 45.5 open at DraftKings and FanDuel to a 48.5 consensus, likely public reaction to Colorado hanging 52 on Weber State. That number is inflated: Colorado's real test was the 14-13 slog at Georgia Tech, and their defense is allowing just 17 ppg. Our totals model lands at 47.8, already below market, and docks 1.55 points for the 7.9 mph wind. The kickoff forecast is rain with 82% precipitation, which trims passing efficiency and can bog down two offenses led by inexperienced quarterbacks. With a strong Colorado defense and wet conditions, I lean under, though the move up despite the weather forecast keeps this a modest lean, not a strong position.

Where our number looks wrong.The EL rating leans on last year and the preseason prior, where Colorado sat at -4.0 with 14% returning production. It cannot see that Colorado's 2026 defense has already held two opponents to 17 ppg including a Power 4 road win, nor that the market has moved a full point-plus toward Colorado since open. It treats a coin-flip as a pass and misses the live signal favoring the dog.

Spread.Our model has Northwestern -4.1, slightly stronger than the market's -3.5 (gap only 0.6, inside the model's own no-edge zone). But the model is mostly prior-driven this early: Northwestern's in-season rating actually fell from 3.315 preseason to 0.689 after beating FCS South Dakota State 34-18 (SDSU scored 18, and NW needed a 24-0 second-half run to pull away), while Colorado's rating climbed from -4.036 to -0.669 after two straight quality results - a road win at Georgia Tech (14-13, on a blocked FG) and a 52-21 blowout of Weber State. Colorado also carries far more blue-chip talent (31.4% vs 4.9%) even with thin returning production (13.7%), suggesting the preseason number undersold a talent gap. The market itself has been unsettled: this line opened between NW -2.5 and -5.5 depending on book, spiked to -4.5 on 9/14, then converged to -3.5 - not a sharp, confident number. True freshman QB Julian Lewis is still erratic (10-21 passing before Colorado's winning GT drive), which caps my conviction. Taking the trend and the talent gap: Colorado +3.5.

Total.The total has climbed from 45.5-47.5 at open to 48-48.5 now, likely priced off two high-scoring wins - but both of Colorado's scoring totals (the 52-21 rout and its share of the 14-13 Georgia Tech win) came against an FCS defense and a low-event nailbiter; the one game against real Power Four competition (Georgia Tech, 9/3/26) went for only 27 total points. Northwestern's own 34-18 win over FCS South Dakota State needed a second-half surge rather than clean execution start to finish. The captured forecast for kickoff (rain, 82% precipitation chance, 7.9 mph wind, dossier 9/16/26) argues down on passing efficiency and tempo for a primetime game. Our totals model, while not a standalone signal at this gap size, also sits a touch under the market (47.8 vs 48.3 consensus). Leaning under 48.5.

Where our number looks wrong.The in-season refit barely differs from the preseason prior after only 1-2 games, so it likely overrates Northwestern off a middling performance against an FCS opponent and underrates Colorado's talent-driven improvement across two real results; injuries, a true freshman QB's growth curve, and the roster overhaul are invisible to it, and the model has no proven ATS edge at any gap size to lean on here.