This was a dark day for the model. A 64.3% straight-up week is barely above season-to-date (62.5% over 64 games), and it arrived amid a slate where the market itself stumbled to an identical 64.3% — meaning the whole week was hard to predict, but the model found no edge against consensus and got caught in the same traps.
The model and market never disagreed this week. That alignment offered no shelter. Both missed all five upsets, including a TB @ DAL collapse where the model's highest-confidence pick — 76% on DAL — came apart completely. DAL was favored by 9.5 and lost by 8, a 17.5-point margin miss that dwarfs almost everything else. Both teams entered depleted: DAL without CB Cobie Durant, LB DeMarvion Overshown, and T Drew Shelton; TB without QB Baker Mayfield, S Antoine Winfield Jr., LB SirVocea Dennis, and CB Benjamin Morrison. The model's read was simply wrong at a high confidence level.
The single worst projection-vs-reality gulf came in CHI @ GB, where the model expected CHI to win by 1 but GB won 31-7. That's a 25-point margin error on a 53% confidence pick — enough to shift the entire week's tone alone. GB's offense executed at a level the model did not anticipate; CHI's defense, without several key pieces (including RB Chris Brooks, WR Savion Williams), collapsed.
Accuracy of 64.3% when the model never strayed from market consensus and the market also hit 64.3% tells a specific story: this was not a week where the model was uniquely overconfident or blind to value. It was a week where the consensus was simply brittle. The model's performance was not up to par, but neither was the market's. That distinction matters for understanding what happened — even if the result is the same regardless.