The model and the market mostly agree in Week 12 — except in six spots where the daylight is real, led by a 4.7-point split on Wake Forest at Georgia Tech.
Fifty-four games carry a market line this week, and across all of them the model and the sportsbooks are separated by an average of just 1.40 points. That's the real story of Week 12: general agreement, not chaos. When a power-ratings model sits that close to closing numbers on most of the slate, it's not failing to find anything — it's confirming the market already did the work.
The widest split on the entire board is 4.7 points, and it belongs to Wake Forest at Georgia Tech. The model has Wake Forest favored by 15.8. The market has Wake Forest favored by 20.5. Both sides agree Wake Forest is the better team by a wide margin — they just disagree on how wide. That's the shape of basically every disagreement this week: not a fight over who wins, but over the width of the margin.
Wake Forest's own power rating, 22.6, ranks seventh nationally, built on an 11.7 offensive number and a defense at -9.4. Those aren't the numbers of a team the model thinks is overrated. The model just isn't willing to go as far as the market on the final score.
Rutgers at Penn State is the second-widest gap on the board, at 4.2 points. The model has Penn State by 21.3; the market has Penn State by 25.5. Same pattern as Wake Forest-Georgia Tech — a market that's more confident in a lopsided result than the model is willing to be.
Arkansas at Texas checks in third at 4.1 points, with the model projecting Texas by 16.4 against a market line of Texas by 20.5. Texas carries a power rating of 22.4, eighth-best in the ratings, anchored by a -13.4 defensive number that's among the stingier marks in the top 12. The model respects that defense. It just doesn't stretch the final margin as far as the books do.
Vanderbilt at Florida and Georgia at South Carolina both land at 3.2-point gaps — Florida favored by 10.3 in the model versus 13.5 in the market, and Georgia favored by 11.3 in the model versus 14.5 in the market. Georgia's overall rating of 31.3 is the best in the country, with a -16.2 defensive number that leads the top 12 by a wide margin. If any team's gap should make you raise an eyebrow, it's the top-rated outfit in the sport still getting shorted by the model relative to the market.
Virginia Tech at Miami rounds out the six at a 3.1-point split — Miami by 17.4 in the model, Miami by 20.5 at the window. Miami's 24.4 rating is third nationally, with the best offensive number in the top 12 at 16.4 points above average.
Zoom out and the six gaps listed here are the exceptions, not the rule. Forty-eight other games on the board have the model and the market closer together than any of these six. A 1.40-point average disagreement across 54 games means the model is, for the most part, drawing the same conclusions the market already priced — which is exactly what you'd expect from a well-calibrated power-ratings system working off the same public information everyone else has.
None of this should be read as a signal to bet one side or the other. The model's own track record makes that explicit: a mean absolute error of 12.1 points against the spread and 12.7 points against the total, over 631 completed games. That's a measure of how close the projections land to final scores — not a betting record, and not evidence of an edge against the closing line. The model is published here for transparency on how it sees the sport, nothing more.