2026 House Odds: Markets 90¢, Models 76-93%

Published 12 October 2026 · Prices frozen 12 October 2026, 16:40 UTC

Every tracker quoting 2026 House odds this week is quoting a stale print. The number they all show is the 92-93% that both exchanges paid in the first week of October; the actual price tonight is 90.5¢ on Polymarket and an 89.5¢ midpoint on Kalshi, three weeks out from a chamber that turns on four seats. That matters more than usual this cycle, because the two best-known public models of the same chamber are 17 points apart — 76% against 93.1% — while their seat projections are under four seats apart. We read both exchanges in the same pass, pulled the full daily history back to the board’s first trade in November 2024, checked Polymarket’s three midterm boards against each other, and read both rulebooks side by side. The headline number is the least interesting thing on the screen.

Where the 2026 House odds stand tonight

There are two liquid contracts on House control and they are priced within a point of each other. Both books are unusually clean: the Democratic and Republican sides sum to exactly 100.0¢ on each venue, which almost never happens on a multi-outcome board.

Kalshi is the deeper venue here, and nobody writes that down. Its House board has traded 53.79m contracts and still has 38.12m open across the two sides, each one settling at a dollar, against $16.53m of lifetime dollar volume on Polymarket’s board. The units are not the same thing — Kalshi reports contracts, Polymarket reports dollars — but there is no reading of those two numbers on which Polymarket is the deeper book. The Senate is where the exchanges actually part company: two full points, 63.5¢ against 61.5¢, on the chamber everyone agrees is close. Live prices for both sit on the House control board and the Senate control board.

Two models, 17 points apart, four seats apart

This is the comparison no odds tracker runs, and it is the only one that tells you whether 90¢ is cheap. Decision Desk HQ’s congressional model put Democrats at 76% for the House on 9 October. The Silver Bulletin’s FLIPR model put them at 93.1% on 8 October. Seventeen points is an enormous gap between two serious forecasts of the same election three weeks out.

Now put the seat projections next to that. DDHQ projects 235 Democratic seats. Silver projects 238.7, against 196.3 Republican. Those are the same forecast: 3.7 seats apart on a 435-seat chamber, both comfortably clear of the 218 needed. The 17-point probability gap is therefore not a disagreement about the election at all — it is a disagreement about how much the projection could be wrong by. One model thinks a 17-seat cushion is nearly safe; the other thinks polling error of the size seen in 2020 and 2022 eats it often enough to matter.

Read down the fourth column and the market’s behaviour is consistent: on the Senate and on the sweep it sits within two points of the two models’ midpoint, and on the House it sits six points above it, nearly on top of Silver. Traders are splitting the difference on the close chamber and siding with the confident model on the one with the cushion.

The market fell three points while the models rose

Kalshi’s contract has been trading since 7 November 2024, two days after the last presidential election, and it opened at 70¢ for the Democrats. It has never printed below 56¢. That is a 23-month-old forecast that has pointed one way the whole time, which is worth holding onto when a tracker presents the current number as news.

The last two rows are the ones to read twice. Polymarket topped out at 93.5¢ on 4 October and has given back three points since; Kalshi gave back 1.9. Over exactly that window the Silver Bulletin’s House number went up, from a 3 September low of 81.1% to 93.1% on 8 October. Markets and model moved in opposite directions for a week, and the trackers quoting “93%” are quoting the 4 October print. The 24-hour move is flat, at 0.0¢.

Polymarket’s own three boards disagree

Polymarket runs three separate contracts that all answer the House question, and they do not give the same answer. Besides the direct board, the Balance of Power board prices all four combinations of the two chambers, and its rungs can be added up to recover each chamber on its own.

Two findings fall out. First, the board is internally inconsistent with its own siblings: it implies 90.00¢ for a Democratic House where the direct board pays 90.5¢, and 62.35¢ for the Senate where the direct board pays 63.5¢. A 1.15¢ discrepancy on a $22.38m book is small in absolute terms and completely ordinary in a market where the two contracts have separate order books — but it means any site quoting “the Polymarket number” is quoting whichever board it happened to read.

Second, and more interesting: the single biggest market-versus-model disagreement in this whole article is not on the House at all. It is the 9.9-point gap on R Senate, D House. The market prices that split at 28.5¢ and Silver’s numbers imply 18.6%. Traders think a Republican Senate survives a Democratic House far more readily than the model does, which is the same bet as being short the sweep.

The Senate ladder says 67.5¢, the market says 63.5¢

There is a fourth way to price Senate control on the same exchange: the Republican Senate seats ladder, eleven rungs covering every outcome from 47 or fewer to 57 or more. With a Republican vice-president holding the tiebreak, Democrats need 51 seats for control, so Democratic control is the sum of the rungs at 49 Republican seats or below.

The raw rungs add to 67.5¢ of Democratic control and the normalised version to 65.06¢ — against 63.5¢ on the direct Senate board and 61.5¢ on Kalshi’s. So the ladder is the most bullish Democratic read on the exchange, the Balance of Power board at 62.35¢ the least, and the spread between Polymarket’s own three answers is 2.7 points wide. The ladder is also the thinnest of the three at $3.14m, so if you are going to trust one, trust it least.

An Iran deal would cost the Democrats the Senate

On 11 October Polymarket opened two conditional boards that price Senate control jointly with a US-Iran agreement over the Strait of Hormuz — one with a 23 October deadline, one with a 2 November deadline, the day before the election. They are the most useful midterm contracts on the exchange and they are 36 hours old.

Both boards say the same thing from different deadlines, which is the reason to believe them: a Hormuz agreement before election day moves the Senate from roughly 64.6¢ Democratic to somewhere between 55.6¢ and 57.4¢. That is a seven-to- nine point foreign-policy dividend, priced explicitly, on a chamber that needs four seats to turn over. The term structure is coherent too: an extra ten days of diplomacy is worth 5.2¢ of additional probability, taking P(deal) from 11.60¢ to 16.80¢.

Treat the conditional arithmetic as indicative rather than exact — it divides one board’s rung by another board’s price, and the 2 November book has traded only $110k. But the direction is unambiguous and it is the cleanest explanation on offer for why the House drifted three points lower in the same week. The underlying diplomacy is tracked on our Strait of Hormuz reopening analysis.

The 2026 House odds price no redistricting penalty

Ten states redrew their congressional maps mid-decade this cycle, and the Cook Political Report puts the net effect at about four extra House seats for Republicans. That is the entire Democratic margin. So the interesting question is whether the market has it in the price — and Polymarket’s popular-vote margin board lets you check.

The modal bucket is a Democratic margin of 8-10%, which brackets the polling average almost exactly: the Silver Bulletin’s generic ballot read D+8.9 on 9 October, D+9.6 on its likely-voter version, and other averages sit at D+7.9 to D+8. So the market believes the polls on votes.

Here is the problem. The board implies a 92.0% chance Democrats win the national House vote, and the control board pays 90.5¢ that they win the chamber. That is a gap of a point and a half. A four-seat structural handicap from new maps should cost more than that in a chamber decided by four seats — it is the difference between DDHQ’s back-of-envelope 222 seats on a uniform 7.5-point swing and Silver’s 238.7. Either the market thinks the redistricting estimate is overstated, or the vote-to-seats conversion is the loosest assumption in the 90.5¢ price. The board is thin enough ($154.5k, double-digit spreads on half its rungs) that this is a flag rather than a finding.

The two venues are not settling the same question

A one-point price difference looks like noise until you read the rulebooks. Polymarket pays on seats; Kalshi’s fallback pays on the Speaker. In a chamber this close that is not a technicality.

The scenario that separates them is specific and not far-fetched: Democrats win 218 or 219 seats and a handful of members withhold votes for Speaker into January. Polymarket has already paid out, because 218 seats is a majority of voting seats. Kalshi may still be open, waiting on who holds the gavel on 1 February. Polymarket’s rules also stop counting any independent who has not announced a caucus by 4 January 2027 — a clause with real teeth in a chamber that currently contains one independent and two vacancies.

Kalshi’s early-determination clause cuts the other way. If the networks call the chamber on election night, Kalshi can settle within days while Polymarket’s contract has already closed to trading at 04:59 UTC on 4 November. Same election, two different instruments. Our Senate flip analysis walks the same comparison through the upper chamber.

What would actually move this before 3 November

Three weeks out, with a 90.5¢ price and a one-cent spread, the House board is not going to move on a poll. The things that could move it are narrower than the coverage suggests.

A Hormuz agreement. Priced at 11.60¢ by 23 October and 16.80¢ by 2 November, and worth seven to nine points of Senate probability against the Democrats. The House board has already moved three points in the week this became live. Watch the conditional boards before the control boards; they will reprice first.

The generic ballot narrowing below D+6. The market assigns 26.0% to a Democratic national margin under six points, and that is the region where a four-seat redistricting handicap starts to bite. Trump’s approval near 32% in the latest Reuters/Ipsos reading is what is holding the current margin up.

Nothing about the model disagreement. DDHQ at 76% and Silver at 93.1% will not converge before election day, because the thing they disagree about — how much polling error to expect — is unobservable until the votes are counted. If you want a view, the question is not which model is right about the House. It is whether 28.5¢ on a Republican Senate alongside a Democratic House is too high or too low.

Related analysis

• Which party will win the House in 2026 — the live board
• 2026 midterms hub — every race we track, with live prices
• 2026 midterms odds: the Senate flipped in 30 days
• Senate control conditional on a Hormuz agreement
• California governor odds 2026: Becerra 96¢, 43 dead rungs
• 2028 presidential election odds: two different favourites

Sources: the Polymarket Gamma and CLOB public APIs (events 32225, 32224, 32228, 106193, 215753, 1171733 and 1172100) and the Kalshi markets, series and candlestick APIs (CONTROLH-2026, CONTROLS-2026), all read at 12 October 2026, 16:40 UTC. Resolution criteria are quoted from the rules published on Gamma and in Kalshi’s own rules fields. Silver Bulletin FLIPR figures for 8 October 2026 — 93.1% House, 74.8% Senate, 74.5% both, 238.7 projected Democratic seats — are as reported by Newsweek from the Silver Bulletin model page, which also publishes the generic-ballot average. Decision Desk HQ’s forecast and its uniform-swing seat arithmetic are from its generic-ballot analysis and its 8 October forecast update. The chamber’s current composition and the seats needed for a majority are from Ballotpedia. The Cook Political Report redistricting estimate, the Reuters/Ipsos approval reading and the historical midterm seat-loss average are from Northeastern Global News, 12 October 2026. Prices here are a frozen snapshot; the live numbers are on the market pages. Nothing here is financial or betting advice.