Going in, our model had favored Baltimore Ravens at 58% win probability, and the pick lost.
Out or doubtful: 2 Ravens, 2 Bills.
PICK: BAL
TIER C
PICK LOSTATS ✗
BAL 58%
BUF 42%
Model confidence: BAL 58%
Our line: BAL -2.7
ODDS
SPREAD
BAL -1.5
MARKET SPREAD · MARKET AVG
BAL -2.7
OUR LINE
MISSED
VS SPREAD
MONEYLINE
BAL · OUR WIN PROBABILITY
58%
BAL · MARKET WIN PROBABILITY
54%
DIFFERENCE
+4%
SOURCE
BAL
BUF
BAL WIN %
Closing line
-130
+110
54.5%
WIN % is each book’s odds as a win chance, with the book’s cut taken out. The card above is their average.
TEAM COMPARISON2025 SEASON · EFFICIENCY, SCORING, TENDENCIES · RANK IN PARENTHESES · FULL-SEASON NUMBERS, NOT AS OF WEEK 1
+0.010Net EPA / play+0.152
+0.034 (12)Offense EPA / play+0.136 (3)
+0.024 (20)Defense EPA / play-0.016 (13)
44.6% (13)Offense success %48.5% (2)
GAME NARRATIVE
BAL C-tier pick misses. BUF takes it.
MISSED
Model picked BAL but BUF won. Projected margin was 2.7, the model overestimated by 3.7 points.
INSTANT CLASSIC
The win probability swung past 50% 2 times. One of the wildest games of the week.
ATS MISS
Missed the spread by 2.5 points.
Every model misses. The edge comes from the aggregate, not any single game.
SITUATIONAL TRENDS
Historical splits, shown for context only. We tested them: picking the situations that beat break-even in training goes 770–760 (50.3%) out of sample, so they are not a betting signal and are not coloured like one.
Every stat on this page is free, and so is the model's pick for this game. Pro is the separate question of which games we would actually bet, and at what stake, plus player props, unit sizing, and the full profit-and-loss ledger.