2026 Midterm Election Odds: Every Race, One Board
The 2026 midterms come down to a single question with four possible answers, and the prediction markets have already put a price on each one. Kalshi's balance-of-power board says the most likely result is a full Democratic takeover of Congress, at 44 cents. We put the same four-way question to a price-blind AI model panel, armed with the complete seat-change record of every postwar midterm, the live polling picture with each pollster's measured accuracy attached, and the actual list of Senate seats in play. The panel agrees with the market about the House. It flatly disagrees about the Senate, and that one disagreement flips the whole board.
The Quick Answer
Kalshi's market and our six-model panel read the House the same way: Democratic, and likely. They split hard on the Senate. The market prices a Democratic Senate near 45 cents; the panel, reasoning from the actual four-seat target list, says 21 percent. That single gap reorders the board and makes a split Congress (Democratic House, Republican Senate) the panel's clear favorite at 55 percent, against a market that ranks the Democratic sweep first. The four-way board, every seat's number, the Senate math behind the call, and a live index of every race we have scored are all below.
Free: The Weekly PM Market Brief — the 8-model panel's graded record, the week's biggest market-vs-model gaps, and what's spiking next. One email, Sundays. Sign up in the box at the end of this article.
New to event contracts? Our guide to how prediction markets work covers the mechanics from zero before you read a single price below.
The Market
| Venue | Kalshi, a CFTC-regulated exchange (18+; availability varies by state, as of August 2026) |
| The Event | "2026 Midterms: Congress Balance of Power" (ticker KXBALANCEPOWERCOMBO-27FEB), four mutually exclusive outcomes, one contract each |
| Settles | On which party controls each chamber as of February 1, 2027, decided by the November 3, 2026 elections |
| Prices As Of | August 8, 2026: D-House/D-Senate 44¢ · D-House/R-Senate 39¢ · R-House/R-Senate 15¢ · R-House/D-Senate 1.4¢ |
This is a neutral, data-driven analysis. It presents both parties' paths identically, advocates for neither, and every electoral fact in it was machine-fetched from the named sources at generation time.
The Board: The Panel's Headline Read
There are exactly four ways November can break. This is the phone-friendly version first, the market price against the panel's blended estimate:
| Outcome | What it means | Market | AI blend |
|---|---|---|---|
| D House / R Senate | Split Congress | 39¢ | 55.2% |
| D House / D Senate | Democratic sweep | 44¢ | 17.5% |
| R House / R Senate | Republican hold | 15¢ | 23.5% |
| R House / D Senate | Reverse split | 1.4¢ | 3.8% |
The single headline: the split that the market ranks second is the panel's runaway favorite, and the sweep that the market ranks first is the panel's largest fade. Here is every seat's number behind that blend:
| Outcome | Market | Fable | Opus | Sonnet | GLM | Kimi | DeepSeek | Blend | Edge |
|---|---|---|---|---|---|---|---|---|---|
| D House / R Senate (Split) | 39¢ | 58 | 55 | 61 | 44 | 58 | 55 | 55.2% | +16.2 |
| D House / D Senate (Sweep) | 44¢ | 20 | 21 | 14 | 8 | 22 | 20 | 17.5% | −26.5 |
| R House / R Senate (R Hold) | 15¢ | 20 | 21 | 21 | 41 | 18 | 20 | 23.5% | +8.5 |
| R House / D Senate (Reverse) | 1.4¢ | 2 | 3 | 4 | 7 | 2 | 5 | 3.8% | +2.4 |
Because the four outcomes describe one election, each model priced the whole partition in a single pass: four numbers from one coherent view of the race, summing to 100, so no model could favor a Democratic House in one contract and contradict itself in another. The chamber odds fall straight out of that math:
| Chamber | Market price | Panel blend | Panel median |
|---|---|---|---|
| Democrats Take The House | 83¢ | 72.7% | 75.5% |
| Democrats Take The Senate | 45¢ | 21.3% | 23% |
Model estimates were generated August 8, 2026, price-blind, from the data card described below. 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. Our full graded record, including the events we have lost to the market, is public.
More live boards from the same panel: House Control 2026 (market: Democrats 83¢) · 2026 Senate control (market: Democrats 45¢) · the full four-way Balance of Power breakdown (market's pick, the sweep, 44¢). Prices fetched August 8, 2026.
What The Panel Was Given
Eight AI models make up the standing panel; six returned verdicts on this run, with the GPT and Gemini seats unavailable and disclosed here. None of them saw Kalshi's prices. Each worked from three fetched layers:
- The historical ledger. In 18 of the 20 postwar midterms the president's party lost House seats, an average of 25.6. The only two exceptions were 1998 (the impeachment-backlash midterm) and 2002 (fought after 9/11 with presidential approval above 60), and each required a dominant national event flipping the usual referendum dynamic. The Senate is structurally stickier, averaging a change of just 3.3 seats, because the map, not the mood, decides the upper chamber.
- The live pulse. As of the last firm reading in the polling feed (late June 2026, roughly six weeks before publication), the generic congressional ballot sat at Democrats +5.0 (45.3 to 40.4 across 56 polls), each poll annotated by its pollster's measured accuracy from 538's graded archive of more than 18,000 polls. Presidential approval stood at 39.7 approve, 57.5 disapprove, a net of −17.8, and midterm losses have historically run worse for presidents under water.
- The volatility band, measured. In presidential years, polls overstated Democrats badly (+3.7 in 2016, +4.7 in 2020). In the Trump-era midterms they were nearly unbiased (−0.3 in 2018, +0.9 in 2022). The models were told exactly that: do not apply a presidential-year correction to a midterm reflexively, and do not assume zero error either.
Where The Panel And The Market Agree
The House. Every layer points the same way and the market already prices it. An out-party House flip is the most reliable pattern in American electoral history (18 of 20 postwar midterms), the generic ballot favors the out-party by 5, and the president's approval is nearly 18 points under water with no 1998 or 2002 style rally condition present. The market says 83 percent, the panel says about 73. The panel is a shade more cautious, citing the 12 weeks still to run, but this is agreement, not an edge.
The reverse split (Republican House, Democratic Senate) at 1 cent. Both sides call it a curiosity. For Democrats to win the Senate while losing the House, the national environment would have to move in opposite directions in the two chambers at once, which is essentially unprecedented in the modern era.
Where They Split: The Senate Is The Whole Argument
The market prices a Democratic Senate near 45 cents. The panel says 21. The models were handed the actual race-by-race landscape, and every one of them made the split their most likely single outcome. Here is the seat math they reasoned from, as the raters currently score it.
Democrats need a net of four seats, and only four Republican-held seats are rated competitive at all:
| Republican-Held Target | Status | Rating |
|---|---|---|
| North Carolina | OPEN (Tillis retiring) | Lean D flip, the best target on the board |
| Maine | Collins, incumbent | Tossup |
| Alaska | Sullivan, incumbent | Tossup |
| Ohio (Special) | Husted, appointed | Tossup |
That is the entire realistic shopping list. The next-best Republican target is rated safe or likely R, and the party's other open seats (Alabama, Iowa, Kentucky, Montana, Wyoming) sit in states it wins by double digits. So "+4" means winning essentially every competitive Republican seat in the country, the lean-flip plus all three tossups. And the defense is not free:
| Democrat-Held Defense | Status | Rating |
|---|---|---|
| Michigan | OPEN (Peters retiring) | Tossup, a live Republican flip chance |
| Georgia | Ossoff, incumbent | Lean D |
| New Hampshire | OPEN (Shaheen retiring) | Lean D |
Lose Michigan, a true coin flip, and the requirement becomes a fifth pickup from the safe-R tier. This is the concrete mechanism behind "a House wave does not carry the Senate." A wave lifts a party's candidates a few points everywhere, which flips dozens of the House's closest districts, but the Senate offers only four flippable seats this cycle, and a few points does not turn Alaska or a Maine incumbent into a safe seat while also securing all three Democratic defenses. It is 2018's lesson running in reverse: that year Democrats gained 41 House seats and still lost two in the Senate, because the map decided the upper chamber.
Where The Panel Changed Its Mind
The panel ran a revision round: after the first pass, every model read the others' anonymized reasoning and could revise. The movement was not cosmetic. Three of the six models started with a Republican hold as their most likely outcome, reasoning heavily from midterm gravity and the difficult map, then flipped to the split after weighing the House base rate their peers leaned on:
- DeepSeek opened with the Republican hold on top, writing that "the House leans Republican due to historical midterm losses for the president's party." After the revision round it reversed the House read entirely: "The House leans Democratic due to the D+5 generic ballot advantage and historical midterm trends favoring the opposition party." Its top outcome moved from the Republican hold to the split.
- Kimi made the same journey, landing on the sharpest one-line summary of the board: "D+5 generic ballot flips House; Senate map forces Dems to run table on four R-held tossups, unlikely." Its Republican-hold price fell from the high 40s to 18.
One model mostly held its ground, and it is worth hearing:
- GLM moved its top outcome to the split but kept a Republican sweep as a loud second, still pricing it at 41 percent against the panel's 23: "The most likely outcome is therefore a split Congress (DR), with a unified Republican Congress (RR) a close second possibility if the low presidential approval proves more decisive than the generic ballot." It is the panel's standing reminder that the environment can still deliver a Republican Senate and a Republican House together.
Every Race That Decides Congress
The balance-of-power board is the headline, but it is built from dozens of individual races, and the panel has scored the ones that actually move it. Each row links to that race's full model verdict. Where Kalshi lists a tradable contract, we show an approximate mark (~) for the Democratic side, fetched August 8, 2026; the individual race books are far thinner than the balance-of-power board and several are untraded, so treat every mark below as directional only and see each race's page for the fuller read.
| Race | Seat | Cook / Sabato rating | Kalshi mark (D side, thin book) | Our verdict |
|---|---|---|---|---|
| North Carolina Senate | OPEN, Tillis retiring | Lean D flip | ~92¢ | Cooper vs. Whatley |
| Maine Senate | Collins (R) incumbent | Tossup | ~64¢ | Collins vs. Jackson |
| Georgia Senate | Ossoff (D) incumbent | Lean D | ~92¢ | Ossoff vs. the field |
| Michigan Senate | OPEN, Peters retiring | Tossup | thin / multi-candidate | El-Sayed vs. Rogers |
| Minnesota Senate | OPEN, Smith retiring | Likely D | ~95¢ | The DFL vs. the field |
| New Hampshire Senate | OPEN, Shaheen retiring | Lean D (Sabato) | untraded | Pappas vs. Sununu |
| Iowa Senate | OPEN, Ernst retiring | Likely R | ~41¢ | Hinson vs. Turek |
| Texas Senate | R-held (Paxton primary) | Lean R | thin, see page | Paxton vs. Talarico |
| Florida Governor | GOP primary | see page | see page | The Republican primary |
| Los Angeles Mayor | Bass (D) incumbent | see page | see page | Bass vs. Raman |
Two chamber-level boards and the deep dives sit alongside the races: House Control 2026, 2026 Senate control, the full Balance of Power four-way breakdown, and, further out, 2028 presidential odds and Trump impeachment odds. If you are new to how these election contracts price and settle, start with our explainer on how election prediction markets work.
Dark Horses The Panel Won't Dismiss
The two splits get the headlines, but the panel keeps real probability on the tails.
- The Republican Hold At 15 Cents, Panel Blend 23.5. This is the panel's quietest edge and its most important dark horse. If the environment reverts toward midterm gravity, or if the polling feed's Democratic generic-ballot lead has faded in the six weeks since its last firm reading, the House flip the whole board rests on becomes shakier, and a Republican House plus a Republican Senate is the outcome that benefits. GLM prices it at 41; even the models that faded it kept it near 20.
- The Reverse Split At 1 Cent, Panel Blend 3.8. Nobody's base case, but the panel would not zero it. A Democratic Senate paired with a Republican House would take a very strange split verdict, some combination of Democrats sweeping the four Senate tossups while the House environment sours late. It is a lottery ticket, and the panel prices it like one.
What Would Change The Panel's Mind
The Senate skepticism rests on a checkable foundation, so specific, event-shaped developments would legitimately move the number:
- A Ratings Change That Expands The Four-Seat Target List. If Iowa or Texas moves from likely-R to tossup, the Democratic Senate path widens and Senate-D rises. Watch the Cook and Sabato updates.
- The Open Michigan Seat Firming Up For Democrats. Michigan is the linchpin defense. A clear Democratic polling lead there removes the "any single loss forces a fifth pickup" problem and lifts both the sweep and Senate-D.
- Special-Election Results Overshooting Partisan Lean. The earliest measurable tell of a true wave, and the same early-warning signal that flagged the Dobbs effect months before the 2022 punditry believed it. A run of specials above lean pushes the House number toward certainty and puts the sweep back in play.
- A Fresh National Reading That Moves Approval Or The Generic Ballot. The current pulse is six weeks stale. A new wave of polls showing approval climbing out of the −18 hole, or the generic-ballot lead shrinking, would soften the House call and lift the Republican hold; the reverse would harden the split.
The Settlement Timeline
| Marker | Date |
|---|---|
| Election Day (All Races Decided) | November 3, 2026 |
| Balance-Of-Power Market Settlement | February 1, 2027 (chambers organized) |
| Next Known Catalysts | Cook/Sabato ratings refreshes through the fall; post-Labor-Day polling wave; any late-cycle retirements or candidate exits |
This page is a standing board. It re-scores when the story moves, and the prices and estimates above carry their own "as of" stamps so you can see exactly how fresh each reading is; last updated August 8, 2026.
How A Contract Pays, And How We Grade Ourselves
To make the prices concrete, and this is mechanics, not a recommendation: every Kalshi contract settles at exactly one dollar if its outcome happens and zero if it does not. At 39 cents, $39 buys 100 contracts of the split; if Congress breaks Democratic-House, Republican-Senate, they settle for $100, a $61 return before Kalshi's trading fees. The panel's 55.2 percent blend implies a fair price near 55 cents for that contract, and the distance between 39 and 55 is the entire argument of this board expressed as one number. The same arithmetic run on the sweep at 44 cents against a 17.5 percent blend is what "overpriced" means in the table above.
You can run any price-against-probability pair like that through the free EV calculator to see what an edge is worth per dollar before you ever place it.
None of this rests on trust. Every number here is stored as a graded row in our public model scoreboard, and when this market settles each model's four-way call gets scored against the actual result, in public, next to every other event we have priced. Every number in this piece gets graded in public once the market settles. This is a forecast made 12 weeks out, and per this series' standing practice, all stored verdicts grade publicly, win or lose.
Election events of this size rarely trade in only one place; Polymarket runs parallel political markets, and when the same outcome is listed on more than one venue, ordinary odds shopping across exchanges sometimes pays an extra cent or two. Liquidity on the balance-of-power board is deep by election-market standards, though the 1-cent reverse split is thin, and thin books move on small money.
To be explicit: these are model estimates for information and entertainment, not financial advice and not a recommendation to trade any contract. This analysis advocates for no party and no outcome. Event contracts are for adults 18+ where legal, on CFTC-regulated venues.
Frequently Asked Questions
Who is favored to control Congress after the 2026 midterms? It depends who you ask. Kalshi's balance-of-power board makes a full Democratic sweep its most likely outcome at 44 cents, with the split Congress close behind at 39. Our price-blind panel disagrees on shape: it makes the split the clear favorite near 55 percent and treats the sweep as the market's biggest overpay at about 18. Both agree the House leans Democratic; they split hard on the Senate.
When are the 2026 midterm elections? Election Day is Tuesday, November 3, 2026. The entire U.S. House is up, along with 33 regular Senate seats plus two special Senate elections (Florida and Ohio), 39 governorships, and thousands of state and local races. The Kalshi balance-of-power contract settles on chamber control as of February 1, 2027.
Why does the AI panel disagree with the market on the Senate? Democrats need a net of four seats, but only four Republican-held seats are rated competitive (North Carolina, Maine, Alaska, and the Ohio special), and Democrats would have to win essentially all of them while also holding tossup Michigan plus Georgia and New Hampshire. The panel judges that near-sweep unlikely even in a friendly year, the same dynamic that gave Democrats a 41-seat House gain and a net Senate loss in 2018.
How does the Kalshi balance-of-power market settle? The board lists one contract for each of the four ways Congress can break. Exactly one settles at a dollar and the rest at zero, based on which party controls the House and the Senate as of February 1, 2027, decided by the November 3, 2026 results.
How were these model estimates produced? Eight AI models make up the standing panel; six returned verdicts on this run (the GPT and Gemini seats were unavailable). Each priced all four outcomes price-blind, working only from fetched base rates, the live polling picture with each pollster's track record, and the 2026 seat map, then read each other's anonymized reasoning in a revision round.
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.
