Houston Texans vs Indianapolis Colts — Win Probability

Model Underestimated Indianapolis in Week 3 Upset Over Houston

The Indianapolis Colts' 19-17 victory over the Houston Texans on Sunday delivered a clear upset, with the model's pregame projection significantly misjudging the outcome.

Before kickoff, the model assigned Indianapolis only a 46% win probability and expected Houston to win by 1.5 points. The market agreed: Houston opened as a 1.5-point favorite. The model's simulation projected a final score of Houston 22, Indianapolis 20.5—a comfortable Texans win.

What actually happened was almost exactly the inverse. Indianapolis won 19-17, flipping the expected margin by 3.5 points (a swing from minus-1.5 to plus-2 in favor of the Colts).

This was a close game masked by an upset label. The actual margin of two points represents a relatively tight contest, not a blowout miss. The model expected Houston to win by roughly a field goal in terms of expected score; the Colts won by a two-point field goal distance. In that sense, the game played out closer to the model's expected competitiveness than the win probability might suggest—it simply went the other direction.

The miss reflects uncertainty baked into the model's pregame assessment. At 46% implied probability, the Colts were given less than a coin-flip chance, but close enough that an upset wasn't shocking. The two-point final margin falls within reasonable variance for a game where the model saw the teams as nearly even in expected output.

For Indianapolis, the victory comes as part of a 1-2 record. For Houston, the loss drops them to 0-3. The model's confidence interval clearly favored Houston, and it was wrong on the winner—but not by an extreme margin in terms of actual points scored.