Denver Broncos at Kansas City Chiefs

Denver heads into Kansas City as the pick across all three model builds: Elo has the Broncos by 3.2 (putting KC's win probability at 39%), Ridge has them by 3.4, and LightGBM is more modest at just 1.7 points. The market sees it differently, favoring the Chiefs at -2.0 with a no-vig win probability of 54%, based on a consensus of nine sportsbooks. That leaves roughly a 5.2-point gap between where the models sit and where the market has settled, with the models consistently leaning more toward Denver. That disagreement is exactly the kind of split the system is built to flag, and both the spread and the moneyline versions of this game qualified as a "Qualified Bet" - the models agree strongly with each other, but the signals around the game are mixed, and a major injury situation adds real risk, with no demonstrated edge over the market established here. Despite qualifying on both sides, only one market type is ever published per game as the official Best Bet, and here that was the moneyline - not the spread. So while the spread analysis met the bar internally, it was not released as a pick; the moneyline was. Worth noting: Ridge and LightGBM's underlying feature outputs couldn't be independently confirmed for this game, the total wasn't sourced, and there's no confirmed final inactives list yet - all reasons to hold the projections a bit loosely until closer to kickoff.

Updated Sep 14, 4:47 PM PDT