What if?

Every number on this site comes from simulating the election thousands of times. Here you choose a scenario, and we keep only the simulations where it happens (or lean on the ones closest to it) and recount. Nothing you do here changes the forecast.

Turn the turnout dial

Suppose one party's voters turn out better than our forecast expects, relative to the other party's. Slide toward the party whose voters show up. The dial converts that into a national House vote and leans on the simulations that land near it. In our model, part of every national swing comes from turnout and part from persuasion: a turnout-driven swing moves close races more than lopsided ones, while persuasion moves every race by about the same amount.

Pick the winners

Choose who wins any race and we keep only the simulations where that happened. Every pick leaves fewer simulations, so stacking many unlikely picks makes the numbers rough; the bar above says how many are left.

How the other races move

If the polls miss like they did in…

Our forecast doesn't assume this year's polls lean either way. Here we rerun it as if every 2026 polling average misses the way its state's final polls did in a past year (pulled toward that year's national average where a state had only a few polls), with the generic ballot missing the same way, and everything else unchanged. These are reruns of the full model, not filtered simulations.