How Sikt's AI Forecast the 2026 Midterms: Every Round, Race and Margin

The 2026 midterms were Sikt Intelligence's first public election benchmark. From September 30 to election morning on November 3, Sikt published 13 rounds of forecasts on control of the House and Senate, ten key Senate races, and how much each race would be won by. Each one carried a timestamp and sat next to Kalshi and Polymarket. Every round is still online, unchanged, and every one will be graded against the certified results.

This page is the record: the final forecast, where Sikt and the markets differed, and a map of every page from the benchmark.

Watch: Sikt's midterm forecast against the markets, why it disagrees in Alaska, and how the S&P 500 might react to each Senate result, in an 11-minute video. The video loads from YouTube when you press play.

The final forecast

The last round was round 13, from Nov 3, 2026, on the morning of election day, before any votes were counted.

Sikt's latest Midterm Bench forecast, from Nov 3, 2026, gives Democrats a 75% chance of controlling the Senate, against 76% on the markets at the time. See every round on Sikt's Senate forecast page.

Sikt's latest Midterm Bench forecast, from Nov 3, 2026, gives Democrats a 91% chance of controlling the House, against 94% on the markets at the time. See every round on Sikt's House forecast page.

Sikt's pick in each key Senate race, next to the market for the same candidate at the time of the forecast:

Race	Sikt's pick	Sikt	Markets then
Iowa	Hinson (R)	51%	50%
Alaska	Peltola (D)	53%	71%
Ohio	Brown (D)	60%	72%
Texas	Talarico (D)	62%	72%
Nebraska	Ricketts (R)	64%	69%
Maine	Jackson (D)	66%	74%
Michigan	El-Sayed (D)	77%	79%
New Hampshire	Pappas	92%	–
Georgia	Ossoff	94%	–
North Carolina	Cooper	96%	–

If every Sikt favorite in the 10 races won, Alaska, Ohio, Texas and Maine would change hands, for 51 Democratic seats and a Democratic majority, assuming the other 25 races go to the party that holds them.

How much each race would be won by

From round 9, Sikt also forecast each race's margin, with an 80% range: the result should land inside it 8 times in 10. In the final round the closest races were Alaska (Peltola +1.0), Iowa (Hinson +1.2) and Ohio (Brown +2.1).

Sikt's middle estimate is 52 Democratic Senate seats, with an 80% range of 48 to 55; 51 is a majority. Kalshi's seat markets put the middle at 52 or fewer (48 to 53+), so Sikt expects the same as the market. See the chart.

Sikt's middle estimate is 230 Democratic House seats, with an 80% range of 219 to 243; 218 is a majority. Kalshi's seat markets put the middle at 241 (217 to 243+), so Sikt expects 11 fewer than the market. See the chart.

For the national House popular vote, Sikt's middle estimate is Democrats by 8.2 points (80% range: Democrats by 4.7 to Democrats by 11.7), against Democrats by 9.8 on Kalshi's market.

Every race, with Sikt's range next to Kalshi's margin market, is in 2026 Senate race margins. The seat counts are in how many seats will Democrats win in 2026?

Where Sikt and the markets differed

Sikt and the markets differ most in Alaska (Mary Peltola: Sikt 53%, markets 71%) and Ohio (Sherrod Brown: Sikt 60%, markets 72%). In the other 5 races the two are within 10 points. In 6 of the 7 races Sikt is less sure of the favorite than the markets are.

Each race page explains Sikt's reasoning in plain language: the history of the state, the polls, the year, and what held Sikt back.

Every page from the benchmark

The midterms hub

2026 midterms: control of Congress, every key race, the Bench table and the margins.
House control forecast and Senate control forecast: Sikt next to the market in every round.
2026 midterm election odds: both chambers, the seat forecasts and the video.

The Senate races

Alaska, Georgia, Iowa, Maine, Michigan
Nebraska, New Hampshire, North Carolina, Ohio, Texas

Each race page has Sikt's read, its margin forecast and a table of every round with the market from the same day.

The articles

Every round as a data file is on the press page.

How the benchmark worked

A forecast every few days. Rounds came every three days from September 30, every two days in the final week, and on election morning.
Its own number. In each round Sikt read the news and data that existed that day and put its own probability on each outcome. Read more in what is an AI superforecaster?
Next to the market. Each forecast was published beside Kalshi and Polymarket, and every comparison on the site uses the market at the time of the forecast, not a later price.
Nothing edited afterwards. Each round shows its date and the version of Sikt that made it.

What happens next

Votes are still being counted and certified, and close races can take weeks. Once the results are certified, every forecast will be scored: the win forecasts against the markets, and the margins against their 80% ranges. The misses count as much as the hits.

The next benchmark runs on SiktBench, starting with a preview of how public AI models forecast real events: which AI is best at predicting the future? Join the waitlist to hear when it starts.

FAQ

Did Sikt Intelligence forecast the 2026 midterms?

Yes. Sikt published 13 timestamped rounds from September 30 to the morning of November 3, 2026, on control of the House and Senate, ten key Senate races and each race's margin, next to Kalshi and Polymarket.

What was Sikt's final midterm forecast?

Sikt's AI forecaster favors Democrats to win both the House (91%) and the Senate (75%). The final round was round 13, from Nov 3, 2026; every race is in the table above.

Where can I see Sikt's forecast for each race?

On each race page, linked above, and on the 2026 midterms page. Each page lists every round with the market price from the same day.

When will Sikt's midterm forecasts be scored?

After the results are certified. Every forecast will be graded, including the ones that missed.