Daily Polymarket prices and poll averages, checked against the results in — Senate and governor races.
Sources: market prices from Polymarket. Polls and results from Wikipedia, used under CC BY-SA 4.0. Not affiliated with or endorsed by Polymarket.
Pick a race on the map. Drag the slider to see what the markets and polls said on any day before the election.
Democrats beat their poll average in most races, by similar amounts. Errors that move together make thirty races worth about two independent observations, so a well-calibrated forecast can still produce a night like this. One election cannot tell polls and markets apart.
On Polymarket, a few hundred wallets put up most of the money. Returns did not rise with wallet size. Small wallets won big where underdog Democrats won and lost where favorites won; the largest wallets mostly stayed near break-even.
Each dot is one size group in one market; zero means the group broke even. The $10,000+ groups hold only a handful of wallets in some races, so single traders can swing them. Hover for counts.
Which races. Every 2026 Senate and governor race with a resolved Polymarket market, a declared winner on Wikipedia, and at least five polls ending in the 30 days before the election as listed on Wikipedia on Nov 2. Races that qualify on polls but are still uncalled are shown and left out of every summary figure.
Market prices. Daily prices from Polymarket's public price-history API, taken at midnight Eastern each day. "Democratic-aligned" means the Democrat, except in Nebraska and Montana, where the main challengers were independents Dan Osborn and Seth Bodnar; there the page uses Polymarket's "Will an independent win?" market. California's governor race was priced per candidate, so it uses the market for the Democratic nominee, Xavier Becerra, starting the day after the June 2 top-two primary. Ohio's and Florida's Senate races are special elections.
Poll averages. Each average uses only the polls Wikipedia listed at the time, rebuilt from the page's revision history (snapshots at the end of each day 1, 7, 14, 21, 30, 45, 60 and 90 days out, then monthly). On a given day, the average takes each pollster's latest poll from the prior 21 days (45 if fewer than three pollsters). A poll that tested several matchups, as most did before the primaries, counts as the average across them, so no average uses hindsight about the nominee. Polls added to Wikipedia after the election are excluded. Polls tagged (R) or (D) on Wikipedia, a partisan lean or sponsor, are drawn hollow; bipartisan teams count as non-partisan.
Poll-implied odds. The poll average converted to a win probability with a normal error of 6.5 points in the final week, rising to 9 points at 90 days out. That spread is an assumption, and the model assumes errors average zero. This year they averaged +3.5 points toward Democrats, so the model was off even though its overall error size was about right.
Results. Vote shares and declared winners from each race's Wikipedia article, read on Nov 5, 2026, with the exact page revision recorded for every race. Shares are of all votes cast. These are unofficial counts and counting continues; where the article gives a share of the count reporting, each race shows it. They will be updated once states certify. Uncalled races are left out of every summary figure.
License. Poll and result figures are adapted from Wikipedia and shared under CC BY-SA 4.0. Market prices are shown for reference only and are not offered for reuse.
Wallets. Polymarket's public trade feed lists the wallet on each side of every fill, makers and takers. The charts count purchases made through 11:59 pm ET on Nov 2 and value each at its purchase price against the final outcome. Later trades are excluded: including them makes large wallets look skilled, because they bought winners after results came in. A wallet is not a person, and one election's returns mostly reflect the shock itself.
About this page. Belvedere Labs is not affiliated with or endorsed by Polymarket. This page is for information only. It is not investment or trading advice, and the figures may contain errors.