Midterm 2026 scorecard: prediction market favourites won 8 of 9 called Senate races, the sweep contract went from 74 to 99 cents, and South Carolina was the miss

Analysis · Updated November 4, 2026 · 18+ · Not legal advice · By BetG8 Editorial · Prediction Markets

Midterm 2026 prediction market scorecard: 8 of 9 favourites won, South Carolina the miss — Nine Senate races called by 11:30 p.m. ET. The favourite won eight. The ninth was priced at 69.5 cents.

Short version: Democrats won both chambers on November 3, 2026, flipping Senate seats in Maine, Iowa, Kansas, Ohio, North Carolina, Texas and South Carolina. Of the nine Senate races called by about 11:30 p.m. ET, the Polymarket favourite at 5 p.m. won eight. The one miss was South Carolina, where Republican Darline Graham was 67 cents on Kalshi and 69.5 cents on Polymarket and lost to Democrat Annie Andrews. The Democratic-sweep contract went from 73.5 cents at 5 p.m. to 98.7 cents at 9 p.m. ET, one to three hours before NBC called the House and Senate. Kalshi reported more than $1.4 billion traded on its midterm markets; Polymarket’s Texas Senate market alone closed with $21.8 million. This is a record of what the prices said and what happened, not a verdict on whether markets beat polls.

LIVE MARKETRepublican Presidential Nominee 2028Trade on Polymarket →

What happened, in the order the markets saw it

Polls closed in the East at 7 p.m. ET. Forbes’ 6:42 p.m. snapshot had the Democratic sweep at 82 cents on Kalshi and 85 cents on Polymarket, the House at 96 and 97, with more than $115 million traded on Kalshi’s balance-of-power market and more than $65 million on Polymarket’s. Two hours later the picture had hardened. The Associated Press called Texas for James Talarico shortly after 9 p.m. ET, the first Democratic Senate win there since 1988 (KSAT). Kalshi’s Talarico contract had been 82 cents at poll close and was 99 cents when the exchange published its own projection, which the company is careful to say “does not mean that races have been called” and does not resolve a market.

The AP called Senate control for Democrats late Tuesday after Josh Turek beat Ashley Hinson in Iowa (Forbes live blog). By the Boston Globe’s count early Wednesday, Democrats had taken Republican-held seats in Maine, where Susan Collins conceded to Troy Jackson, North Carolina (Roy Cooper), South Carolina (Annie Andrews), Texas, Ohio (Sherrod Brown over Jon Husted), Kansas (Adam Hamilton over Roger Marshall) and Iowa, and had flipped at least 12 House seats with more to come (Boston Globe). Alaska, which uses ranked-choice voting, was not called at the time of writing. The live prices for the national contracts remain on our 2026 midterm election odds page; at 7:30 UTC on November 4 Polymarket’s “Democratic Party controls the Senate” contract was 99.95 cents and still open, because the contract resolves on the seated majority rather than on a network call.

The scorecard: nine races, one miss

The cleanest test is the set of Senate races that had been called by about 11:30 p.m. ET, scored against the last Polymarket price before 5 p.m. ET, as compiled by Cryptoticker from Polymarket’s data and NBC’s calls. Kalshi’s noon November 2 prices, from our Election Day article, are shown where we have them.

Race	Favourite	Polymarket, 5 p.m. ET Nov 3	Kalshi, noon Nov 2	Winner	Favourite won?
South Carolina	Darline Graham (R)	69.5¢	67¢	Annie Andrews (D)	No
Iowa	Josh Turek (D)	50.5¢	52¢	Turek	Yes
Kansas	Adam Hamilton (D)	64.5¢	63¢	Hamilton	Yes
Ohio	Sherrod Brown (D)	72.5¢	71¢	Brown	Yes
Texas	James Talarico (D)	76.5¢	71¢ (Nov 3, Fortune)	Talarico	Yes
Florida	Ashley Moody (R)	87.5¢	–	Moody, 51-48	Yes
New Hampshire	Chris Pappas (D)	95.0¢	–	Pappas	Yes
Georgia	Jon Ossoff (D)	97.2¢	–	Ossoff, 64-36	Yes
North Carolina	Roy Cooper (D)	98.0¢	–	Cooper	Yes

Eight of nine is the headline, but it is the wrong number to stop on. The average favourite in those nine races was priced at 79 cents. If the prices were honest probabilities, the expected number of upsets in nine races is about 1.9, and the observed number was one. A well-calibrated market loses some favourites; a market that never loses one is overpricing them. On this sample the prices were, if anything, slightly too cautious, not too confident. The Brier score across the nine, which penalises confident wrong answers heavily, works out to 0.11 against 0.25 for a coin flip and zero for perfect foresight. Nine races is a small sample and we say so again below.

South Carolina: what a 30-cent winner looks like

The miss deserves the detail. Lindsey Graham died on July 11, 2026 and Darline Graham, endorsed by President Trump on July 23, was priced at 85 to 87 cents through the summer, according to the price history in Kalshi’s projection note. Andrews was 15 cents on October 14. A New York Times/Siena poll then had the race tied at 48, Kalshi’s own polling average had Graham up about 1.5 points, and the contract moved to 67-33 on November 2. At poll close Andrews was 39 cents. Kalshi projected her win at 3:09 UTC with the contract at 99 cents. On Polymarket, Andrews was 30.5 cents at 5 p.m. ET and the South Carolina Senate market resolved to the Democrats with $4.04 million traded, of which $2.2 million was on the Democratic side.

Two things are worth separating. A 30-cent outcome happening is not evidence that the price was wrong; it is what a 30-cent outcome is supposed to do roughly three times in ten. What would be evidence is a pattern, and the pattern on the night ran the other way: the only real toss-up, Iowa at 50.5 cents, landed on the favourite, and every race priced above 70 cents went the way the price said. The second point is about speed. A 39-cent contract going to 99 cents in the hours after polls close is not the market changing its mind; it is the market watching the count. Markets price information faster than networks publish calls, and they are allowed to be wrong while doing it, which is why Kalshi’s rules settle on media calls and Polymarket’s on the seated majority rather than on their own prices.

How the sweep contract moved against the clock

The national contracts tell the timing story better than any single race. Polymarket’s both-chambers contract, which had opened Election Day at 76.5 cents, was 73.5 cents at 5 p.m. ET, 88.5 at 7 p.m., 92.5 at 8 p.m. and 98.7 at 9 p.m. ET, per Cryptoticker’s hourly reads. NBC called the House at 10:08 p.m. ET and the Senate at 11:26 p.m. ET. In other words, the market had moved from three-in-four to near-certain between one and three hours before the two network calls that Kalshi’s chamber contracts use for early settlement.

Time (ET, Nov 3)	Polymarket Democratic sweep	Event
~2 a.m.	76.5¢	Our Election Day read of the API
5 p.m.	73.5¢	Last pre-results price
6:30 p.m.	85¢ (Kalshi 82¢)	Forbes snapshot
7 p.m.	88.5¢	Eastern polls close
8 p.m.	92.5¢	–
9 p.m.	98.7¢	AP calls Texas shortly after 9
10:08 p.m.	–	NBC calls the House
11:26 p.m.	–	NBC calls the Senate

The pre-results price is the one to remember. At 73.5 cents on Polymarket and 74 cents on Kalshi the day before, a Democratic sweep was a clear favourite and nothing like a certainty; Forbes noted that Kalshi’s sweep contract had been below 40 cents before the Iran war began. The result landed on the favoured side of a one-in-four risk, which is consistent with the price and proves nothing about it on its own.

The money

Fortune reported, from data Kalshi supplied on November 3, that trading on the exchange’s midterm markets had passed $1.4 billion, with $251 million on control of Congress, more than $181 million on the House, more than $112 million on the four-way chamber combination market, just under $110 million on the Los Angeles mayoral race and close to $100 million on the Texas Senate race (Fortune). Polymarket’s figures, as cited by Cryptoticker, were about $164 million across its three national markets and $67 million across 14 Senate markets. Our own read of Polymarket’s public API at 7:30 UTC on November 4 showed the Texas Senate market closed at $21.8 million, the Senate control market at $31.2 million and still open, and South Carolina at $4.0 million.

The gap between the two venues is the point. Kalshi’s Texas market was roughly four and a half times the size of Polymarket’s, and its Congress contract alone exceeded Polymarket’s entire national book. For a reader deciding where a price is most informative, volume is not accuracy, but it is a reasonable proxy for how much it costs to move the number. On the night, the two venues disagreed by one to three cents on every race we could compare, which is inside the spread.

The letter nobody answered

November 3 was also the deadline Senators Elizabeth Warren and John Curtis gave Kalshi and Polymarket to explain how they prevent insider trading on election contracts, which we covered when the letter went out. Gaming America reported on Election Day that neither company had published any of the four figures the senators asked for (Gaming America). The letter has no legal force, and a private reply to the senators would not show up in public. But the first question after a night in which the markets were broadly right is going to be who was trading in them, not whether they were right, and the companies have not yet answered it in public.

Method and limits

Race-level Polymarket prices are the last trade before 5 p.m. ET on November 3 as compiled by Cryptoticker from Polymarket data; Kalshi prices are from Kalshi’s November 2 noon note and its two election-night projection posts, and the Talarico figure of 71 cents is from Fortune on November 3. Results and call times are from the AP as reported by KSAT, Forbes and the Boston Globe, and from NBC via Cryptoticker for the chamber calls in Central European Time, converted to ET. Volume figures are the companies’ own, reported by Fortune and Cryptoticker, except the three Polymarket markets we read directly from the Gamma API at 7:30 UTC on November 4, which refer to the international exchange, not Polymarket US. Nine races is a small sample: a Brier score of 0.11 on nine binary events has wide error bars, and a single additional miss would have moved it to about 0.16. Alaska, Michigan, Nebraska and most House seats were uncalled when we wrote this and are excluded. This article does not predict anything and is not advice to trade.

Frequently asked questions

How accurate were Kalshi and Polymarket in the 2026 midterms?

In the nine Senate races called by about 11:30 p.m. ET on November 3, 2026, the Polymarket favourite at 5 p.m. ET won eight. The average favourite was priced at 79 cents, which implies about 1.9 misses in nine races; there was one. The Brier score across the nine was 0.11 against 0.25 for a coin flip.

Which race did the prediction markets get wrong?

South Carolina. Republican Darline Graham was 67 cents on Kalshi at noon on November 2 and 69.5 cents on Polymarket at 5 p.m. ET on November 3. Democrat Annie Andrews won. Kalshi had Andrews at 39 cents when polls closed and 99 cents when it projected her win.

When did the Democratic sweep contract reprice?

Polymarket’s both-chambers contract went from 73.5 cents at 5 p.m. ET to 88.5 at 7 p.m., 92.5 at 8 p.m. and 98.7 at 9 p.m. ET. NBC called the House at 10:08 p.m. and the Senate at 11:26 p.m. ET, so most of the move came one to three hours before the network calls.

How much was traded on the 2026 midterm prediction markets?

Kalshi reported more than $1.4 billion on its midterm markets by November 3, with $251 million on control of Congress and $181 million on the House. Polymarket reported about $164 million across its three national markets and $67 million across 14 Senate markets; its Texas Senate market had $21.8 million when it resolved.

Did markets call races before the networks?

Prices moved before calls, but a price is not a call. Kalshi’s Talarico contract was 82 cents at poll close and 99 cents when Kalshi published its projection; the AP called Texas shortly after 9 p.m. ET. Kalshi’s own note says a projection does not mean a race has been called or a market resolved.