Washington Commanders vs Dallas Cowboys — Win Probability

Dallas dominates projection in 37-20 rout of Washington

The model called Dallas to win, but not like this. Before kickoff, the projection system gave the Cowboys a 62% win probability with an expected final score of 27.2–23.3—a margin of just 3.9 points. The market line of Dallas minus-4 was actually wider than the model's own expected margin, suggesting the betting public saw a slightly more decisive outcome than the simulation did.

What happened on the field was a far more comfortable victory. Dallas won 37–20, a 17-point margin that more than quadrupled the projected spread. The Cowboys' actual score (37) exceeded the projection by nearly 10 points, while Washington's output (20) fell 3.3 points short of the expected 23.3.

This represents a significant miss in both the magnitude and nature of the game. The model nailed the winner but substantially underestimated Dallas's dominance. A 3.9-point spread projection that became a 17-point blowout is a gap of 13.1 points—the kind of error that suggests the pregame read on team strength, tempo, or execution was materially off.

The accuracy on win probability (62% for the eventual winner) held up, since Dallas did win. But for a model whose brand depends on precise margin projection, missing by 13 points is a clear miscalibration of how much better Dallas would perform. The Cowboys' performance exceeded both the projection and the market line in the same direction, indicating neither model nor market fully captured Dallas's edge heading into the game.

This was a miss dressed as a hit—correct pick, wrong confidence level and wrong expected result.