Atlanta Falcons vs Green Bay Packers — Win Probability

Atlanta's 35-14 Win Over Green Bay Was a Major Upset; Model Favored Packers Heavily

The Atlanta Falcons stunned Green Bay 35-14 on Sunday, delivering one of the week's clearest departures from pregame expectation. The model assigned Atlanta only a 37% win probability before kickoff, reflecting confidence in a Packers victory. The market agreed, installing Green Bay as a 4.5-point favorite.

The model's expected final score was Green Bay 23.5, Atlanta 19—a Packers win by 4.5 points. That projection tracked closely with the betting line, suggesting alignment between the model's simulation and public consensus. Both were decisively wrong.

Instead, Atlanta scored 35 points and held Green Bay to 14, winning by 21. The Falcons' actual margin exceeded the model's expected Packers margin by 25.5 points—a massive gap that swings the entire game outcome and magnitude in Atlanta's favor.

This was a genuine upset. The Falcons' 37% pregame win probability means the model viewed this as a clear underdog situation, yet they won by three scores. The 21-point actual margin versus the -4.5 expected margin (favoring Green Bay) represents a swing of 25.5 points from projection to result.

When a model assigns a team a 37% win probability and that team wins decisively, it signals either that the pregame inputs—team strength ratings, recent performance data, or other factors—substantially misjudged one of these rosters, or that single-game variance exceeded what the simulation anticipated. A 21-point victory is not within normal simulation noise.

Both teams fell to 1-2 after the game, leaving their overall 2026 records equal despite dramatically different Week 3 outcomes. The model's read on this matchup was among the week's most inaccurate.