The Marginbeta
NFL · Super Bowl odds + bracketSnapshot date unavailable

NFL Super Bowl odds & projected playoff bracket

Elo + QB · 50K Monte Carlo · Market shrink

Every team in the league sorted by the model's probability of winning the next Super Bowl. The model regresses last year's ratings toward an offseason mean (new QB, new coach, big roster moves). The bracket is played out 50,000 times; the headline title odds are gently shrunk toward the sportsbook consensus to stay honest about the model's edge.

Loading projections…

How the model works

Each team starts the offseason with an Elo rating carried over from the end of last season, then regressed toward the league mean to account for roster turnover. Coaching changes and quarterback moves get explicit Elo adjustments from the QB registry, since a new starter can swing a team's outlook more than anything else on the depth chart.

From those ratings the model plays the entire season and postseason forward 50,000 times using a 538-style win-probability formula (home-field worth about 55 Elo points). The bracket tables are the empirical rates across those simulations.

For the headline title odds, the raw simulation is shrunk toward the sportsbook's implied probability whenever the two disagree by more than two percentage points. That keeps the published edge honest: the model's pure conviction is shown in the “Pure sim” column, while the headline number blends it with the market consensus.