How accurate are Polymarket election odds?

The 2026 Senate and governor races, 67 of 68 resolved on Polymarket

The 34 U.S. Senate races and 34 governor races Polymarket lists with party markets, one market each, all 68 priced by every cutoff. Polymarket lists only candidate markets for Alaska’s two races and California’s governor race, so those three are not graded. As of 2026-11-12, 67 of 68 races had resolved on Polymarket. Its market on the Kansas governor race was still open, and that race stays out of the tables until its market resolves. The first cutoff, midnight Eastern going into Election Day, is the election-morning cutoff in the tables below.

Of the 68 graded 2026 races, 67 had resolved on Polymarket by 2026-11-12. In those 67, the party priced above 50 cents at midnight Eastern going into Election Day won 64. The party priced above 50 cents seven days out won 62 of them, and thirty days out 61. In the 67 of 68 graded 2026 races resolved by 2026-11-12, Polymarket’s prices had a Brier score (the average squared gap between price and outcome: 0 is perfect, and always pricing 50 cents scores 0.25) of 0.0430 on election morning, 0.0553 seven days out and 0.0668 thirty days out. In 46 of those 67 the election-morning favorite was priced at 90 cents or more, which pulls the score toward 0. These scores cover races only and are not comparable with the 2024 Brier scores further down, which score election markets (545 of 688 priced on election morning), not races. Before Polymarket’s market on the Mississippi Senate race resolved, this page’s election-morning Brier score was 0.0324 on 66 races. Adding that race, where the favorite was priced at 86.0 cents and lost, moved it to 0.0430 on 67. Among the 67 resolved races, the three election-morning favorites that lost were the Republican sides of the Texas governor race (60.2 cents), the South Carolina Senate race (68.0 cents) and the Mississippi Senate race (86.0 cents). Three races changed favorite between the thirty-day and election-morning cutoffs, and all three resolved to the later favorite. The Republican side of the Georgia governor race moved from 52 to 30 cents, the Iowa Senate race from 57 to 49 cents, and the Kansas Senate race from 67 to 38 cents.

Scope: One market per race: the Republican-side market for each of the 34 U.S. Senate races and 34 governor races Polymarket lists with party markets for the 2026 midterms (Alaska Senate, Alaska governor, California governor list candidate markets only and are not graded). Each is scored on its last traded price at midnight Eastern going into Election Day, and again seven and thirty days earlier. The favorite is whichever party was priced above 50 cents at that moment. This measures Polymarket prices, not any trader; a race with no trade by a cutoff is absent from that cutoff, and the markets on 1 of the 68 races were still open at the time of the check. A band with fewer than 30 races is shown but never quoted.

Price freshness: Each cutoff price is the last trade at or before that moment, not a trade made at it. At midnight Eastern going into Election Day, 0 of 68 races had last traded more than seven days before the cutoff. Seven days out, 3 of 68 had (Wyoming governor, Massachusetts Senate, Rhode Island Senate). Thirty days out, 10 of 68 had. The oldest priced trade was 37 days before its cutoff. None of them was priced between 10 and 90 cents.

2026 Senate and House control

The two chamber-control markets, priced at the same three cutoffs. Each row shows the market’s question and, in parentheses, the side whose price is shown, which on both markets is Yes, the Republican side.

The Republican side of the Senate control market traded at 35.0 cents thirty and seven days out and 24.0 cents going into Election Day. The Republican side of the House control market traded at 7.0, 9.0 and 6.0 cents at the same cutoffs. Both resolved No, so Republicans won control of neither chamber in the 2026 midterms.

The 2026 races by how confident the price was

As of 2026-11-12, 67 of 68 races had resolved on Polymarket. In 46 of those 67, the favorite was priced at 90 cents or more on election morning, and every one of those favorites won; the favorites in that band averaged 97.0 cents, and the 95% interval on the win rate runs from 92% to 100%. The other four bands hold between one and nine races each, too few to quote.

Why do 2026 midterm accuracy figures differ?

This page and PredictDesks’ race-by-race scorecard (dated November 6) both grade Polymarket’s 2026 midterm prices, and they report different counts of how often the favorite won: 64 of 67 races here, 63 of 65 markets there. Each counts a different set of markets and takes the price at a different moment.

PredictDesks’ 65 markets are 63 Senate and governor races, all of them among this page’s 67, plus the two chamber-control markets. On the 63 races both grade, the two scorecards name the same favorite and the same winner. On those 63 shared races Polymarket’s favorite won 61, and the two favorites that lost among them were the Republican sides of the Texas governor race and the South Carolina Senate race. This page counts four races PredictDesks does not include. The favorite won three of them and lost the Mississippi Senate race, which takes 61 of 63 to 64 of 67 here. PredictDesks adds the two chamber-control markets, which this page shows in their own table above. The favorite won both, which takes 61 of 63 to 63 of 65 there.

The Brier scores also cover different markets at different moments. This page’s 0.0430 covers its 67 races priced at midnight Eastern going into Election Day, and PredictDesks’ 0.0318 for Polymarket (its summary box rounds it to 0.032) covers its 65 markets at 6 p.m., so the gap between them does not show which prices were more accurate.

PredictDesks’ scorecard excludes the Mississippi Senate race, which was uncalled when it published, and says the race stays out until it settles. Polymarket’s market on that race has since resolved, and this page now counts it. The Republican side of Polymarket’s Mississippi Senate market was priced at 86.0 cents on election morning and lost. Neither scorecard includes the Kansas governor race. Polymarket’s market on it was still open as of 2026-11-12, and this page adds it when it resolves.

Kalshi’s own preliminary working paper, “Kalshi and the 2026 Midterms: Calibration and Accuracy” (November 2026), reports that the candidate priced above 50 cents won 98.2% of races at poll close. That figure grades Kalshi’s prices on 510 races called by 6:26 a.m. Eastern on November 4, most of them House races, each priced at its last trade before the polls closed in its state. The paper does not score Polymarket. This page grades no individual House races and prices each race at midnight going into Election Day, hours before any poll closed, so 98.2% and this page’s 64 of 67 do not show which venue priced the midterms better.

The 2024 general election, 94 races

Fifty state presidential contests, 33 U.S. Senate races and 11 governor races in OVERROUND’s record, one market each, all 94 priced by every cutoff. The first cutoff, midnight Eastern going into Election Day, is the election-morning cutoff in the tables below.

The party priced above 50 cents at midnight going into Election Day won 91 of the 94 races. The party priced above 50 cents seven days out won 93 of the 94, and the party priced above 50 cents thirty days out won 88. Three races changed favorite between the thirty-day and election-morning cutoffs, and all three resolved to the later favorite. The Republican side of the Nevada presidential contest moved from 47 to 62 cents, the New Hampshire governor race from 49 to 70 cents, and the Ohio Senate race from 48 to 60 cents.

Scope: One market per race: the Republican-side market for each of the 50 state presidential contests, 33 U.S. Senate races and 11 governor races in OVERROUND's record from the 2024 general election. Each is scored on its last traded price at midnight Eastern going into Election Day, and again seven and thirty days earlier. The favorite is whichever party was priced above 50 cents at that moment. This measures Polymarket prices, not any trader; a race with no trade by a cutoff is absent from that cutoff. A band with fewer than 30 races is shown but never quoted.

Price freshness: Each cutoff price is the last trade at or before that moment, not a trade made at it. At midnight Eastern going into Election Day, 1 of 94 races had last traded more than seven days before the cutoff (Delaware Governor). Seven days out, 0 of 94 had. Thirty days out, 1 of 94 had (Washington Governor). The oldest priced trade was 12 days before its cutoff. None of them was priced between 10 and 90 cents.

The presidency and both chambers

The three headline markets, priced at the same three cutoffs. Each row shows the market’s question and, in parentheses, the side whose price is shown. That is the side Polymarket lists first in the market, which is Yes on a Yes/No market and the first-named party on a control market.

The Trump market traded at 50.4 cents thirty days out, 66.2 cents seven days out and 59.3 cents going into Election Day. The House control market traded at 50.0 cents on its Democratic side at that last cutoff and resolved Republican.

Races by how confident the price was

Seventy-three of the 94 races had a favorite priced at 90 cents or more on election morning, and every one of those favorites won; the favorites in that band averaged 97.6 cents and the 95% interval on the win rate runs from 95% to 100%. The other four bands hold between one and eight races each, too few to quote, and the table shows them so the composition is visible. With 94 races, most bands are thin by construction; in 2024, 73 of them landed in the top band.

The 688 two-outcome “US Election” markets resolved November 5, 2024 to January 24, 2025

The 94 races sit inside a wider set of 688 two-outcome markets Polymarket tagged “US Election” that resolved between November 5, 2024 and January 24, 2025, with voided and disputed markets left out. Polymarket’s own sub-tags for them include State Presidential Election, Senate, Swing States, Popular Vote, Margin of Victory, Mentions and Election Night. This table scores each market’s first-listed side at each cutoff. Of the 143 markets without a price on election morning, 142 first traded after the cutoff (post-election questions) and one never traded.

Each row is its own set of markets, so a difference between rows is partly a difference in which markets were priced; the rows do not show accuracy changing as Election Day approached. Brier is the average squared gap between price and outcome, where 0.25 is what always pricing 50 cents scores. Accuracy is how often the side above 50 cents won, with markets at exactly 50 cents left out of that rate. The table below sorts the 545 election-morning markets by price.

The table does not rise steadily through the middle. The 30 to 40 cent bin resolved 18.2%, the 50 to 60 cent bin 11.1% and the 70 to 80 cent bin 57.1%, each below the bins on either side. The six bins from 30 to 90 cents hold nine to seventeen markets each, too few for one election to pin a rate. The two ends, where 292 and 87 markets sit, resolved 1.0% and 100.0% against average prices of 1.7% and 97.4%.

The 2026 races on Polymarket

As of 2026-09-22, Polymarket listed 1,177 open events under its midterms tag, carrying 11,606 markets. Its race tags listed 35 U.S. Senate race-winner events and 36 state governor race-winner events; California’s governor race is filed under the governor tag, not the midterms tag, and lieutenant governor and territory races are not counted. Each link opens the race on Polymarket. The 2026 table at the top of this page grades the races Polymarket had resolved by 2026-11-12. The Kansas governor race is added when its market resolves, and Alaska’s two races and California’s governor race are not graded.

Scope of the 2026-09-22 listing: Counts the race-winner events Polymarket lists under its midterms, governor and Senate race tags at the time of the check (35 U.S. Senate races and 36 governor races), each carrying its own party or candidate markets. Polymarket adds races and markets through the cycle, so the count is a same-day snapshot. No price is quoted for an open race; the graded table is published after the election resolves.

How to read these tables

A calibrated price is one where favorites priced at 90 cents win about 90% of the time. The stated price is the average price in the band; the win rate is how often the favorite actually won; the interval is the range that many races can pin the true rate to (a 95% Wilson interval). Fixed calendar cutoffs replace the benchmark’s distance-from-resolution horizons here because election markets resolve days or weeks after the vote, as races are called. A cutoff price is the last trade at or before that moment, not a trade made at it. In a thinly traded safe seat that trade can be days or weeks old, and each election’s price freshness line says how many races that applies to. One market per race avoids double counting. In a two-party race the Democratic market’s price is a dollar minus the Republican market’s, so scoring both would count each race twice. The same scoring, on distance-from-resolution horizons and across all of Polymarket’s markets, is the calibration benchmark; the football season graded the same way is on the NFL and college football page; how wallets rather than prices are graded is on the methodology page.

How accurate were Polymarket's odds in the 2026 midterm elections?

As of 2026-11-12, 67 of the 68 Senate and governor races in OVERROUND's record had resolved on Polymarket; its market on the Kansas governor race was still open. The 68 are the 34 Senate and 34 governor races Polymarket lists with party markets (Alaska's two races and California's governor race have candidate markets only). In the 67 resolved races, the party priced above 50 cents at midnight Eastern going into Election Day won 64, a rate whose 95% interval (the range of underlying rates consistent with 64 wins in 67 races) runs from 88% to 98%. Seven days out the favorite won 62 of 67, and thirty days out 61 of 67. In the 67 of 68 graded 2026 races resolved by 2026-11-12, Polymarket's prices had a Brier score (the average squared gap between price and outcome: 0 is perfect, and always pricing 50 cents scores 0.25) of 0.0430 on election morning, 0.0553 seven days out and 0.0668 thirty days out. In 46 of those 67 the election-morning favorite was priced at 90 cents or more, which pulls the score toward 0. Among the 67 resolved races, the three election-morning favorites that lost were the Republican sides of the Texas governor race (60.2 cents), the South Carolina Senate race (68.0 cents) and the Mississippi Senate race (86.0 cents). By race type, the election-morning favorite won 32 of the 34 Senate races and 32 of the 33 governor races resolved on Polymarket. Going into Election Day the Republican side traded at 24.0 cents in the Senate control market and 6.0 cents in the House control market, and both resolved No. These are Polymarket's prices, not any trader's record.

How accurate were Polymarket's odds in the 2024 election?

Across the 94 races in OVERROUND's record (50 state presidential contests, 33 U.S. Senate races and 11 governor races), the party priced above 50 cents at midnight Eastern going into Election Day went on to win 91 of them, a rate whose 95% interval runs from 91% to 99%. Seven days out the favorite won 93 of 94, and thirty days out 88 of 94. By race type, the election-morning favorite won 48 of 50 state presidential contests, 32 of 33 Senate races and all 11 governor races. Across the wider set of 688 two-outcome election markets on this page, 545 carried a price at that moment, and among those, the side above 50 cents won 92.5% of the time (markets at exactly 50 cents are left out of that rate). These are Polymarket's prices, not any trader's record.

What did Polymarket price Trump at before the 2024 election?

The market on Donald Trump winning the presidency traded at 50.4 cents thirty days before the election, 66.2 cents seven days out, and 59.3 cents at midnight Eastern going into Election Day, and resolved Yes. The market on Democrats controlling the Senate traded at 19.0 cents at that last cutoff and resolved to Republicans. The market on Democrats controlling the House traded at 50.0 cents at that cutoff and resolved to Republicans.

Does Polymarket have markets on the 2026 midterm elections?

Yes. As of 2026-09-22, Polymarket listed 1,177 open events under its midterms tag, carrying 11,606 markets. Its race tags listed 35 U.S. Senate race-winner events and 36 state governor race-winner events (California's governor race is filed under the governor tag, not the midterms tag); they are listed on this page with links. This page grades the 68 races that carry a Republican-side party market (34 Senate, 34 governor) as Polymarket resolves them, using the same method as the 2024 table. Alaska's Senate and governor races and California's governor race list candidate markets only and are not graded. No price is quoted for a race whose market is still open.