New England Patriots vs Buffalo Bills — Win Probability

Patriots Stun Bills: Upset Win Defies 7-Point Projection Gap

New England entered Sunday as a 30-percent proposition to beat Buffalo, according to the pregame model. The market agreed, installing Buffalo as a 7-point favorite. The simulation projected a Bills win by 7 points, with Buffalo scoring 28.3 and New England 21.2.

What happened instead: New England won 29-26, a 3-point victory that inverted the expected outcome entirely.

This was an upset by definition—the model favored Buffalo at 70 percent probability—and the margin tells the full story of the miss. The pregame projection called for Buffalo to win by 7. New England won by 3. That 10-point swing between expected and actual result represents a substantial gap, though not catastrophic. The Patriots outscored their projection by 7.8 points; the Bills fell 2.3 short of theirs.

The model's scoring projections were directionally reasonable—both teams landed within striking distance of their pregame estimates. New England's 29 points beat the projected 21.2, a meaningful but not historic outperformance. Buffalo's 26 came close to the 28.3 projection, a near-miss that didn't account for the full collapse in the final outcome.

What this gap means: The model identified Buffalo as the stronger team and did so with clear confidence. The 7-point spread and 70-percent win probability were not fence-sitting. New England's victory suggests the simulation either undervalued the Patriots' offensive or defensive capacity, or Buffalo's performance deteriorated beyond what the pregame data suggested. A 10-point margin gap between projection and result is the kind of miss that warrants review of the inputs, though single-game upsets, especially in small samples, are part of any model's reality.