2026 midterms on prediction markets: what the odds said in October

If you want to know what prediction markets think about control of Congress after 3 November 2026, the answer in mid-October was clear on one chamber and less clear on the other. Traders priced a Democratic House at roughly 91% to 93% and a Democratic Senate at roughly 62% to 65%, with the Senate number well below what at least one poll-based forecast model was showing. This post sticks to what was published on or before 22 October 2026, and it says nothing about results.

What were Kalshi and Polymarket pricing in October?

A DeFi Rate market tracker updated on 15 October 2026 listed three Democratic contracts on each exchange. Democrats to win the Senate stood at 62% on Kalshi (with $41M traded) and 62% on Polymarket ($5.5M traded). Democrats to win the House stood at 91% on Kalshi ($71M traded) and 93% on Polymarket ($13M traded). A Democratic sweep of both chambers was 61% on Kalshi and 63% on Polymarket.

The same tracker's log shows an entry on 21 October 2026 with Democratic Senate control at 65% on Kalshi and 64.5% on Polymarket, and the House at 91%. So the Senate drifted up about 3 points in under a week, while the House barely moved. For context, a Kalshi News article published on 30 September 2026 put Democratic Senate control at 63%, a Democratic sweep at 62%, a Democratic House with a Republican Senate at 28% and a Republican sweep at 8%. A 62% Senate price means traders gave Republicans about 4 chances in 10 of holding the chamber, as how market odds and resolution work explains.

How did that compare with polls and forecast models?

Newsweek's James Bickerton reported Silver Bulletin model figures with a latest data point of 8 October 2026. They gave Democrats 74.8% to win the Senate (an average of 52.2 seats) and 93.1% to win the House.

Set side by side, the House agreed: about 93% in the model, 91% to 93% in the markets. The Senate did not. The model was at about 75% on 8 October, while Kalshi and Polymarket were at 62% on 15 October and about 65% on 21 October. The dates differ by a week or more, so this is a rough comparison, not a measured gap.

Polls measure something different again. An AP-NORC poll fielded 8 to 12 October 2026 of 925 registered voters, reported by the Associated Press on 15 October, found about half of registered voters preferring Democratic control and roughly 4 in 10 preferring Republican control, with 15% undecided. That is a national preference, not a count of seats. For more on how the two approaches differ, see prediction markets vs. polls.

How did markets do in 2022 and 2024?

Mixed, and it depends on what you count. In 2022, the Deseret News reported on 1 November 2022 that PredictIt had Republicans at 73 cents for the Senate and 90 cents for the House, on seven-day running averages. The same article noted that the final New York Times polls gave Democrats a slight edge to keep the Senate, the opposite of the market. CoinDesk reported on 27 October 2022 that Polymarket gave Republicans an 87% chance of winning the House and 64% of taking control of Congress overall.

Afterward, Devin O'Connor wrote for Casino.org on 13 November 2022 that Democrats had kept the Senate with 50 seats, and that Republicans were expected to take the House. So the House call held up while the Senate call, in markets that leaned Republican, did not.

For 2024, the most careful public work I found is a paper by Joshua Clinton and TzuFeng Huang, dated 1 December 2025. It examined more than 2,500 political markets on the Iowa Electronic Markets, Kalshi, PredictIt and Polymarket in the final five weeks of the 2024 presidential campaign. PredictIt was the most accurate, with 93% of its markets correctly predicting outcomes better than chance, while Kalshi and Polymarket were less so. The authors also found identical contracts priced differently across exchanges. Our longer take is in are prediction markets accurate.

Why is "who controls the Senate" not one single bet?

Wording matters more than most people expect. Kalshi's own explainer, published 16 October 2026, says its Senate control market settles on a consensus of media organizations about which party controls the Senate. If there is no consensus, it settles on the party of the Senate president pro tempore on 1 February 2027. The article describes this as the contract's current rules, which means they could change.

The sweep contracts add another layer. DeFi Rate described them as effectively Senate-control trades with an added House requirement. That is why the Democratic sweep sat at 61% on Kalshi, one point under the 62% Senate price: it is close to the same bet. If you hold both, you are not diversified.

A market that can settle as late as February 2027 ties up your position far longer than one that settles on the election result. Always read the rules text before you trade, and see prediction market resolution: who decides for how disputes get handled.

What about individual state races?

The chamber markets are the deepest ones. On 15 October the House contract on Kalshi had $71M traded and the Senate contract $41M, while Polymarket's Senate contract had $5.5M. A single state race has far less money behind it, so a modest order can move the price several points.

I did not find verified, dated state-race prices from before 22 October that I can quote, so I will not quote any. The general point stands anyway. The Clinton and Huang paper found cross-exchange price gaps even in the most heavily traded 2024 markets, and thinner markets have more room for that. Check the volume figure before you read anything into a 5 point move.

How should you use these prices?

Use them as a second opinion next to models and polls, not as a replacement. When the three agree, as they did on the House, confidence is reasonable. When they disagree, as on the Senate in mid-October, the honest reading is that nobody knew, and the price was a range of opinion.

Prediction markets also cannot tell you why a price is where it is. If you want to test your own view without risking anything, you can make a prediction on any of these questions with free credits and see how your reasoning holds up against the crowd.

Ravioli is a free-to-play prediction market. You trade with Sense, which is free, and the prizes are real.

Create a free account and browse the board.