Republican House Seats 2026: How Many Will Republicans Hold?
Most 2026 midterm coverage asks one binary question: does the party in power keep the House or lose it? Kalshi's seat-count board asks the sharper one. Instead of a single yes-or-no, it splits the outcome into eleven mutually exclusive buckets, from Republicans finishing below 193 seats to the 218-222 range to above 237, and prices each one. That turns "who wins the House" into a full probability distribution over the exact size of the result. We fetched the live board, hid the prices from a six-model AI panel, and asked each model to build that distribution from scratch.
The Quick Answer
As of August 4, 2026, Kalshi's market implies Republicans are roughly a 1-in-5 bet to keep the House. The buckets at 218 seats or more add up to about 20¢, so the market has Democrats near 80% to flip the chamber. Our six-model AI panel is a little more bearish on Republicans still, pricing the GOP's chance of holding at 16%. Where do they agree? Both center the most likely outcome on Republicans winning roughly 203-207 seats, a loss of around a dozen-plus off their current majority, and a modest Democratic House. Where they split is the tails: the market's single biggest bucket is a Republican collapse below 193 seats at 18¢, and the panel calls that overpriced at under 7%. The full bucket-by-bucket board, every model's number, and where the panel breaks from the money are below.
📩 Get the settled verdict. We grade every call on this board in public once the 2026 results are certified. Want the scorecard, plus a re-scored board if the polls move this fall? Grab it from the email box at the bottom of this page. Nothing here is a prediction of fact.
New to event markets on elections? Start with our plain-English guide to betting sports and events on Kalshi before you read a single cent as an edge.
Prices on this page refresh during active news waves; every quote below carries an "as of" stamp so you can see when it was last read.
The Market
| Venue | Kalshi, a CFTC-regulated event-contracts exchange (18+; availability varies by state, see where Kalshi is legal) |
| Event Ticker | KXRHOUSESEATS-27 |
| The Contract | Eleven mutually exclusive markets, one per seat-count band. Exactly one resolves YES: the band containing the number of U.S. House seats the Republican Party holds in the 120th Congress |
| Settlement | Resolves February 1, 2027, on the Republican Party's actual House seat total per the certified results of the November 3, 2026 election |
| Majority Line | 218 of 435 seats is a majority. Bands of 218-222 and above mean Republicans keep the House; 217 or fewer means Democrats take it |
| Closes | On resolution; expected expiration February 1, 2027 |
Verify it yourself: the live board, its exact bands, and current YES prices sit on Kalshi under event ticker KXRHOUSESEATS-27. Current House composition and race ratings are public at Ballotpedia's 2026 House elections page and the Cook Political Report House ratings.
The Board
Here is the full board in phone-friendly form. "Market (YES)" is the live Kalshi quote; "AI panel blend" is the equal-weight mean of the six models after a revision round. Prices as of August 4, 2026. These are model estimates, not predictions of fact and not financial or trading advice.
| GOP Finishes With… | What it would mean | Market (YES) | AI panel blend |
|---|---|---|---|
| Below 193 | Democratic landslide | 18¢ | 7% |
| 193-197 | Large Democratic wave | 10¢ | 12% |
| 198-202 | Solid Democratic majority | 15¢ | 18% |
| 203-207 | Modest Democratic majority | 15¢ | 20% |
| 208-212 | Narrow Democratic majority | 13¢ | 16% |
| 213-217 | Democrats flip by a hair | 12¢ | 11% |
| 218-222 | Republicans hold by a thread | 8¢ | 8% |
| 223-227 | Republicans hold outright | 5¢ | 4% |
| 228-232 | Republicans expand | 3¢ | 2% |
| 233-237 | Republican wave | 2¢ | 1% |
| Above 237 | Big Republican wave | 4¢ | <1% |
| 218 Or More (Majority) | Republicans keep the House | ~20¢ | 16% |
The shape is the story. Both columns lean the same way (a Democratic House is the base case for the market and the panel alike), but they are shaped differently. The market's mass is spread wide and carries a fat left tail: its single largest bucket is a Republican collapse below 193 seats at 18¢. The panel's mass is a tighter hill centered on 203-207, and it prices that deep-collapse tail at less than 7%. One column is pricing a real chance of a blowout in either direction; the other is pricing a fairly ordinary midterm loss.
Every Seat's Number
The full panel, band by band. These are the round-two numbers, after each model read the others' anonymized reasoning and was free to revise. Columns may not sum to exactly 100 because of rounding.
| GOP Band | Fable | Opus | Sonnet | GLM | Kimi | DeepSeek | Blend |
|---|---|---|---|---|---|---|---|
| Below 193 | 6% | 6% | 8% | 6% | 8% | 6% | 7% |
| 193-197 | 10% | 11% | 11% | 11% | 13% | 16% | 12% |
| 198-202 | 17% | 18% | 15% | 18% | 18% | 24% | 18% |
| 203-207 | 20% | 21% | 16% | 20% | 20% | 22% | 20% |
| 208-212 | 18% | 18% | 15% | 16% | 16% | 12% | 16% |
| 213-217 | 13% | 12% | 13% | 12% | 11% | 7% | 11% |
| 218-222 | 8% | 7% | 10% | 8% | 7% | 6% | 8% |
| 223-227 | 4% | 4% | 6% | 4% | 4% | 4% | 4% |
| 228-232 | 2% | 2% | 3% | 2% | 2% | 2% | 2% |
| 233-237 | 1% | 1% | 2% | 1% | 1% | 1% | 1% |
| Above 237 | 1% | 1% | 1% | 1% | <1% | <1% | <1% |
| Majority (218+) | 16% | 15% | 22% | 16% | 14% | 13% | 16% |
Model estimates generated August 4, 2026, from a price-blind data card: each model saw the live historical base rates, the current generic-ballot and approval polling, and the Cook race-ratings battlefield, but never the market's prices. This run reached six of the eight seats we normally poll; the ChatGPT and Gemini seats were unreachable on this machine (an expired login and a missing local client), and their absence is disclosed here rather than hidden. These are model estimates, not predictions of fact and not financial or trading advice. Models are frequently wrong, and the market price reflects real traders' money.
Where The Panel Splits From The Money
Start with what everyone agrees on, because it is the real headline: a Democratic House is the base case. The market prices Republicans keeping their majority at about 20%, the panel at 16%, and both center the single most likely result on Republicans winning roughly 203-207 seats, a loss of about 13 to 17 off the seats they hold now, which would hand Democrats a majority around 230. The expected seat count implied by both columns lands near 206 for the GOP. On the center of the distribution, the models and the money barely disagree.
The disagreement lives in the tails, and it runs in one clear direction: the market prices bigger blowouts than the panel does. The clearest example is the deep-collapse bucket. Kalshi's largest single bucket is Republicans finishing below 193 seats at 18¢, a Democratic landslide of 243-plus seats. The panel prices that same outcome at under 7%. Kimi K3 put the reasoning for a tighter distribution plainly:
Mode stays 203-207; mean ~206, implying a ~14-seat GOP loss. — Kimi K3
The panel's case is not that the macro climate is friendly to Republicans; it plainly is not. It is that the battlefield structure caps how large a loss that climate can produce. As Claude Opus reconciled the two:
Macro climate points to Democratic gains, but Cook rates more D-held tossups (10) than R-held (8) and nearly every competitive seat shifted right 2020-2024, cushioning losses below the -25.6 base rate. — Claude Opus
That is the whole split in one sentence. The market reads a hostile national environment and keeps real weight on a wave that runs the score up; the panel reads the same environment through a district map that it thinks limits the damage, and squeezes its mass toward a normal-sized Democratic win.
The Bear Case For Republicans (Why A Democratic House Is Favored)
The case that Republicans lose the House is the base case, and it rests on three fetched facts that all point the same way. First, history: since 1946, the president's party has lost House seats in 18 of 20 midterms, by an average of about 26 seats. Second, the current climate: the generic congressional ballot has averaged Democrats +5 to +6 in recent polling, and President Trump's approval sits deeply underwater at roughly 40% approve to 58% disapprove, a net near -18. Third, the exception test: the only two postwar midterms where the president's party gained House seats were 1998 and 2002, and both required presidential approval near or above 60%, a condition nowhere in sight today. DeepSeek folded all three into its read:
The mode remains centered on a GOP seat loss of roughly 15-22 seats (198-207), consistent with a D+5-6 generic ballot and -18 approval. — DeepSeek
GLM 5.2 added the historical yardstick that makes the loss concrete: the last time a Republican president faced numbers like these, the House swung hard.
A D+5-6 generic ballot in 2018 produced -41 seats, but that was against a less right-shifted map. — GLM 5.2
If you want the binary version of this same question (simply, does the GOP keep the House or not), our House control 2026 verdict prices that single market head-to-head against the panel.
The Bull Case For Republicans (Why The Loss Stays Contained)
The reason the panel does not price a blowout is structural, and it is the most important number on the board that is not a poll. According to the Cook Political Report's June 2026 House ratings, there are 18 toss-up seats, but 10 of them are held by Democrats and only 8 by Republicans, and Cook notes that nearly every competitive district in the country shifted toward Republicans between 2020 and 2024. In plain terms: even a national environment that favors Democrats has to fight a battlefield where Democrats are defending more of the truly up-for-grabs seats. Claude Sonnet leaned on exactly that to justify keeping more weight on a Republican hold than the rest of the panel:
Cook rates more Democratic-held toss-ups than Republican-held, and the map shifted right 2020-2024. — Claude Sonnet
The other pillar of the bull case is arithmetic on the current majority. Republicans start from a razor-thin edge of roughly 220 seats, so the distance to hold is short: they can lose a handful of seats and still clear 218. Every model priced a real, if minority, chance that the losses stay small enough. But none mistook "contained" for "likely to hold": as Kimi K3 noted, holding the majority "requires near-zero losses, implausible at -18 approval and D+5-6 with no 1998/2002-style rally condition present."
Dark Horses The Panel Won't Dismiss
The tails are thin, but they are not dead, and they run both ways:
- Republicans Hold (218+). The panel blend is 16%, but the spread is real: Claude Sonnet priced the majority at 22%, well above the rest, arguing the toss-up map is a bigger cushion than the pack credited. If the battlefield tilt matters more than the national mood, the way it did for Senate Democrats in 2018 who lost seats even in a wave year, the GOP-hold bands are where that shows up. A confirmed shift in the generic ballot back toward the center this fall would reprice this whole right side of the board fast.
- A Deep Democratic Wave (Below 193). This is the market's favorite tail at 18¢ and the panel's biggest skeptic at under 7%, so it is the single sharpest disagreement on the board. The path to it is a late polling miss in the other direction from 2016 and 2020, a Democratic-underestimate year like 2018, when a Republican president's party shed 41 seats. Every model kept a few points of mass here precisely because the historical loss distribution is so wide (the worst postwar midterm was -63). Fable was explicit that the fat left tail earns its weight:
The historical loss distribution is wide (worst -63), so I moved a point of mass toward the 193-202 range rather than trimming it. — Claude Fable
The Convergence Round
We run these panels twice. After the first pass, each model reads the others' anonymized distributions and may revise. The revisions here were small but truly two-directional, which is the sign of a healthy round rather than a herd. DeepSeek opened as the most bearish seat on the Republican majority at 6% and, after peers stressed the right-shifted battlefield, raised it to 13%, shifting mass out of the deepest-loss bands and into the middle. Sonnet moved the other way on the same evidence, trimming its outlier 29% majority read down to 22% but declining to fully converge, and GLM slid its peak one band to the right, from 198-202 to 203-207. Nobody abandoned a defended position to conform; the blend that came out, 16% on the majority with a clean peak at 203-207, is tighter and better argued than the first pass, and it still sits four points under the market on the question that matters most.
How To Read This Board Like A Sharp
A prediction-market price is a probability with a dollar sign in front of it. The 18¢ on "below 193" is the market's shorthand for "about an 18% chance," and because a venue like Kalshi trades close to no-vig, the eleven prices are a cleaner read on true probabilities than a lopsided sportsbook line would be, though on this board they sum to a little over 100¢, which is the exchange's spread, not an edge you can bank. If turning cents into probabilities is not yet second nature, our guide to what a prediction-market price actually means walks through it.
The move a sharp makes next is not to trade the market's number, it is to compare it against an independent estimate and act only when the two diverge. That is exactly what this panel is for: six models priced the board blind, from the real base rates and current polling rather than a headline. On this board that second opinion agrees with the market on the center, a modest Democratic House, and disagrees on the shape, calling the market's fat collapse tail too rich. The Senate half of the 2026 map tells its own story; our Michigan Senate verdict breaks down one of the seats that could decide the other chamber.
Every number in this piece gets graded in public once the market settles, so you can check whether the panel or the market read the 2026 House better. You can follow the running tally on our full graded scoreboard.
To be explicit: nothing here is advice to buy or sell anything, and nothing here is an endorsement of any party or candidate. Kalshi is a CFTC-regulated event-contracts exchange, you must be 18 or older to use it where it is offered, and midterm forecasts made 13 weeks out move as the results come in. These are model estimates, not predictions of fact and not financial or trading advice, and the market price reflects real money from traders who may know more than any model does.
Hottest Prediction Markets Right Now
- 2028 Democratic presidential nominee · $166M traded
- Rocket Classic Winner · $124M traded
- 2027 Pro Football Champion · $51M traded
- Pro Baseball Champion · $40M traded
- Bitcoin price at the end of 2026 · $30M traded
- 2027 Pro Basketball Champion · $19M traded
Every market above links to our full AI model verdict; browse them all on the OddsShopper prediction markets hub, and see how every settled call actually scored on the full graded scoreboard.
