The model's widest gap between its number and the market lands on a Chiefs-Dolphins under, and the model wants no part of 44.5.
Kansas City at Miami is posted at 44.5. The model has it at 41.8. That's a 2.7-point gap, and it's not a rounding error — it's the largest separation between model and market anywhere on this week's totals board. When the model disagrees with Vegas by that much, the disagreement itself is the story.
The market spread on this same game sits at -8.5, which tells you the number 44.5 was built with a lopsided game in mind. The model's projection of 41.8 is its own read on total scoring, arrived at independently of how the market split the spread. Two different numbers pointing the same direction — under — is worth noting without pretending to know why.
This play carries a Grade A, and an A here sits in the middle of the model's grading range — not the top, not the bottom. It's a strong signal, not a lock, and it shouldn't be read as top-tier conviction just because the letter is high in the alphabet. The 8.48% edge attached to it is real, but real and guaranteed are different words with different meanings.
That distinction matters more on the play with the biggest gap on the board, because size of gap and size of grade aren't the same measurement. This one just happens to have both working in its favor relative to everything else the model spit out this week.
Four other totals plays carry grades this week, all Bs. Las Vegas at New Orleans is an under at 45.5 with a 7.93% edge. Arizona at San Francisco is an over at 47.5, 7.61% edge. Philadelphia at Chicago is also an over at 47.5, 6.93%. Tennessee at New York Giants rounds it out as an under at 43.5, 5.09%.
Across all nine of the model's totals plays this week, the average gap between model number and market number is 1.7 points. The smallest gap is 0.9, the largest is the 2.7 sitting on KC-Miami. Four of the nine plays are overs, five are unders — a mixed board, not a themed one. That split doesn't tell you anything about how these games will unfold on the field. It just tells you the model isn't leaning one direction across the whole slate.
This is a model's output on a limited sample, run through this week's slate and nothing more. There's no established track record being cited here, and none should be implied. An 8.48% edge and a Grade A are descriptive of the model's own math, not a promise about Sunday.
Treat it as analysis, not a financial instruction. The gap is the largest the board has produced this week, the grade lands in the middle of the model's own scale, and both of those facts are worth knowing before anyone treats 44.5 like a settled number.