2026 Midterm Election Odds: Every Race On One Board

Updated August 12, 2026 · 13 min read · by Dave Loughran

Kalshi prices the 2026 midterms three ways at once: a four-way board on how Congress breaks, a market on each chamber, and a separate contract on all 35 Senate seats being contested. Put those together and you get a single picture of what money thinks November looks like. We put the same four-way question to eight AI models, price-blind. seven of the panel's eight seats returned a board. They agree with the market almost exactly about the House. They are 20 points apart from it on the Senate, and that one disagreement is the whole story of this cycle.

The Quick Answer

The market and the panel read the House the same way: Democratic, and close to settled. Kalshi prices a Democratic House at 85¢; the panel blends to 85%. The Senate is where they part company. The market prices a Democratic Senate at 50¢. The panel, reasoning from the actual list of seats in play, says 30%. That gap makes the split outcome, a Democratic House against a Republican Senate, the panel's clear favorite at 58% against a 39¢ price. The four-way board, the seat-by-seat math, every Senate race with a live price, and the specific things that would change the panel's mind are all below.

New to this kind of market? How election prediction markets work explains what a price actually means before you read one.

The Market

Venue	Kalshi, a CFTC-regulated exchange (18+; availability varies by state, as of August 2026)
The events	"2026 Midterms: Congress Balance of Power?" (four mutually exclusive outcomes), plus separate House control, Senate control, and per-state Senate winner contracts
Settles	On which party controls each chamber as of February 1, 2027, following the November 3, 2026 elections
Prices as of	August 12, 2026: R-House/R-Senate 15¢ · R-House/D-Senate 1¢ · D-House/R-Senate 39¢ · D-House/D-Senate 48¢
Note on quotes	Long-dated election contracts often show no live bid or ask between trades. Prices here are the most recent traded level, read at build time and stamped above.

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

Outcome	What it means	Market	AI blend	Gap
D House / R Senate	the split	39¢	58%	+19
D House / D Senate	a Democratic sweep	48¢	28%	-20
R House / R Senate	Republicans hold both	15¢	13%	-2
R House / D Senate	the reverse split	1¢	2%	+1

The chambers, on their own:

Chamber	Market	Panel blend	Panel range
Democrats take the House	85¢	85%	83 to 86
Democrats take the Senate	50¢	30%	25 to 32

These chamber numbers are not separate estimates. Each model priced all four joint outcomes in one pass, and the chamber odds are read back out of that single board, which is why nothing here can contradict itself.

More live boards from the same panel: Senate control, race by race (Republicans 50¢) · House control (Democrats 85¢) · North Carolina, the cycle's one rated flip (91¢ Democratic). Prices fetched August 12, 2026.

Every Seat's Number

Outcome	Claude Fable	Claude Opus	Claude Sonnet	GPT	GLM	Kimi	DeepSeek	Blend
D House / R Senate	56	58	57	56	58	63	55	58%
D House / D Senate	29	28	26	30	27	23	30	28%
R House / R Senate	13	12	15	12	13	12	13	13%
R House / D Senate	2	2	2	2	2	2	2	2%
Democratic House (derived)	85	86	83	86	85	86	85	85%
Democratic Senate (derived)	31	30	28	32	29	25	32	30%

Every seat on this panel is graded against real market settlements — records to date: Claude Fable 82% on 460 graded calls · Claude Opus 84% on 524 graded calls · Claude Sonnet 83% on 508 graded calls · GPT 83% on 5,138 graded calls · GLM 81% on 2,318 graded calls · Kimi 83% on 2,296 graded calls · DeepSeek 80% on 2,343 graded calls. Recomputed daily; the full scoreboard is public.

Model estimates generated August 12, 2026, price-blind, from the data card described below, after a revision round in which each model read the others' anonymized reasoning. These are model estimates, not predictions of fact and not financial or trading advice. Models are frequently wrong; the market price reflects real traders' money.

Every number above is stored and graded against the actual result when this market settles. Our running record on every event we have priced, wins and losses, sits on the full graded scoreboard.

Eight seats were asked and seven returned a board within the run window. The blend is the equal-weight mean of those seven.

What the Panel Was Given

Four layers, every one machine-fetched at run time. No prices appeared anywhere in the material the models saw.

1. The historical ledger. All 20 postwar midterms. The president's party lost House seats in 18 of 20, averaging a loss of 25.6 seats. The two exceptions each needed a dominant national event: 1998, the impeachment-backlash midterm, and 2002, fought after September 11 with presidential approval still above 60. The Senate is structurally stickier, averaging a change of 3.3 seats, and the reason is the map rather than the mood. The clearest demonstration is 2018, when Democrats flipped 41 House seats and still lost two Senate seats, because they were defending 26 of the 35 seats up, ten of them in states Trump had carried.
2. The national frontier. The generic congressional ballot averaged across six public aggregators, and presidential approval averaged across ten, each read from the current published tables rather than a poll feed that might have gone quiet. Those numbers appear in the next section.
3. The live seat map. Every one of the 35 Senate seats being contested, with its Cook Partisan Voting Index, whether the incumbent is retiring, and each of 11 handicappers' current rating, split into seats a clear majority of raters call competitive and seats where only one or two do.
4. The measured error band. In presidential years, polls overstated Democrats by 3.7 points in 2016 and 4.7 in 2020. In the two Trump-era midterms they were nearly unbiased: 0.3 the other way in 2018, 0.9 in 2022. The models were told not to apply a presidential-year correction reflexively and not to assume zero error either.

The National Environment

Two numbers drive the House side of this board, and both are near the edges of their historical range.

The generic congressional ballot averages Democrats +6.7 across six public aggregators, and the striking part is the agreement: every one of them lands between +6.5 and +7.0. Aggregators using different pollster weights and different smoothing rarely cluster that tightly, and several models flagged that as evidence the signal is real rather than an artifact of one house's method.

Presidential approval averages 37.8 approve against 59.3 disapprove, a net of 21.5 points under water, across ten aggregators. That matters mechanically rather than rhetorically: the historical link between a president under water and his party's midterm House losses is one of the sturdiest patterns in the ledger, and it is also the number that rules the 1998 and 2002 exceptions out. Both of those years needed approval far above water or a rally event. Neither condition is present.

Kalshi lists a contract on the generic ballot itself. The current week's market asks whether VoteHub's time-weighted Democratic margin will sit above 6.7% when it settles, and it last traded at 72¢. That is worth understanding for what it is: the strike sits almost exactly on the current average, which is why it trades near a coin flip. It is a market on a weekly wiggle, not on November.

That environment cashes out as a seat count — our Republican House seats model verdict prices the exact ladder.

Where the Panel Splits From the Market

The House: agreement, and unusually precise agreement. The market says 85¢. The panel says 85%. Seven independent models, none of which saw the price, landed within a point of it. There is no edge here, and the reason the panel is worth reading on this board is that it earns the right to disagree elsewhere by matching the market where the market is obviously right.

The Senate: a 20-point gap. The market prices a Democratic Senate at 50¢. The panel says 30%, and every seat on the panel landed below the market. The argument is arithmetic rather than atmosphere. Democrats need a net of four seats. Working from the live ratings, here is the shopping list they would be working from, with the current Kalshi price on the Democratic side of each race.

Republican-held target	PVI	Rating spread	Democratic price
North Carolina	R+1	Lean D flip/Lean D flip/Likely D flip	91¢
Maine	D+4	Tossup/Tossup/Tossup	66¢
Ohio (special)	R+5	Tossup/Tossup/Tossup	51¢
Alaska	R+6	Tossup/Tossup/Tossup	52¢
Iowa	R+6	Lean R/Lean R/Tossup	40¢
Texas	R+6	Lean R/Lean R/Tossup	53¢

Democratic-held defense	PVI	Rating spread	Democratic price
Michigan (open seat)	EVEN	Tossup/Tossup/Tossup	56¢
Georgia	R+1	Lean D/Likely D/Likely D	91¢
New Hampshire (open seat)	D+2	Lean D/Lean D/Likely D	89¢

North Carolina is the one seat every rater scores as a Democratic flip. After that the path runs through genuine coin flips in states that lean Republican by 5 or 6 points, and it runs through them while Michigan, an open seat rated a tossup by every rater on the board, has to be held first. Reaching a net of four means winning North Carolina, then three of Maine, Ohio and Alaska, and dropping none of the three defenses. Lose Michigan and the requirement becomes a fifth pickup from Iowa or Texas, both of which sit in states that voted Republican by six.

That is the mechanism behind a sentence people say about every midterm without meaning anything specific by it: a House wave does not carry the Senate. A few points of national swing flips dozens of the House's closest districts, because the House offers dozens of close districts. The Senate this cycle offers about four, and a few points does not turn Alaska into a safe seat while simultaneously securing an open tossup in Michigan.

Where the Panel Changed Its Mind

Each model priced the board once, then read the other forecasters' reasoning with the names stripped off and was invited to revise. Revision was optional. The Senate number moved on six of the seven seats, and it converged hard: the opening spread of 19 points closed to 7.

The biggest move came from the most bullish seat. DeepSeek opened at 41% on a Democratic Senate and came down to 32:

"The 2018 mechanism, same electorate, opposite chamber results, remains central: a wave sufficient to flip the House may still fall short in the Senate." — DeepSeek

The two most bearish seats moved the other way, which is what a real convergence looks like rather than a stampede. Claude Opus opened at 22% and revised up to 30, and its reasoning names the exact seat that sets the ceiling:

"Sweeping three tossups while holding an even-PVI open seat lands near 30%. Chambers are strongly correlated but asymmetric." — Claude Opus

One seat did not move at all. GPT priced a Democratic Senate at 32% in both rounds, and explained the hold rather than defending it:

"Iowa and Texas provide upside, but both sit in R+6 states and are still Lean R/Tossup rather than core Democratic targets." — GPT

What Would Change the Panel's Mind

The panel's Senate skepticism rests on checkable things, which means specific, dated events would legitimately move it. These are the ones the models named:

• The generic ballot tightening below Democrats +3, or widening past +9. Aggregators update daily. Tightening pushes probability toward Republicans holding both chambers; widening pushes it toward a Democratic sweep as the tossups start breaking together.
• Michigan getting reclassified. It is currently a tossup on every rater's board and it is the single largest swing variable in the Senate math. Moving to Lean Democratic raises the sweep; moving to Lean Republican forces Democrats to find a fifth pickup and pushes the split higher.
• Iowa or Texas moving to tossup. Both are Lean Republican now. Either one becoming a genuine tossup widens the target list from four seats to five and is the cleanest single upgrade available to the Democratic Senate case.
• Presidential approval recovering past roughly 44, or a rally-type national event. This is the 1998 and 2002 mechanism. It is the only development on this list that weakens the base rate the entire board rests on.
• The final district lines. Several states used revised maps from the 2025 and 2026 mid-decade redistricting round, with expected gains on both sides. Claude Opus specifically held its Republican-holds-both number above the rest of the panel on this basis, and it is the one card fact that can blunt a national margin in the House.

Every Senate Race On The Board

All 34 contested Senate races Kalshi lists, with the current price on each side and where the handicappers have the seat. Prices fetched August 12, 2026; long-dated contracts trade thinly, so treat a one-cent move as noise rather than news.

State	Leader	Price	Other side	Read
Alabama	Barry Moore	98¢	Everett Wess 2¢	Safe
Alaska	Democratic party	52¢	Republican party 48¢	Battleground
Arkansas	Tom Cotton	96¢	Hallie Shoffner 6¢	Safe
Colorado	John Hickenlooper	97¢	Mark Baisley 3¢	Safe
Delaware	Democratic party	98¢	Republican party 2¢	Safe
Florida	Republican party	85¢	Democratic party 14¢	Watch list
Georgia	Jon Ossoff	91¢	Mike Collins 9¢	Competitive
Idaho	Jim Risch	94¢	Todd Achilles 5¢ · David Roth 2¢	Safe
Illinois	Juliana Stratton	98¢	Don Tracy 2¢	Safe
Iowa	Ashley Hinson	60¢	Josh Turek 40¢	Battleground
Kansas	Roger Marshall	79¢	Adam Hamilton 20¢	Watch list
Kentucky	Andy Barr	94¢	Charles Booker 7¢	Safe
Maine	Troy Jackson	66¢	Susan Collins 35¢	Battleground
Massachusetts	Democratic party	99¢	Republican party 1¢	Safe
Michigan	Abdul El-Sayed	56¢	Mike Rogers 43¢	Battleground
Minnesota	Peggy Flanagan	93¢	Michele Tafoya 8¢	Safe
Mississippi	Cindy Hyde-Smith	94¢	Scott Colom 8¢	Safe
Montana	Kurt Alme	90¢	Seth Bodnar 10¢ · Alani Bankhead 2¢	Watch list
Nebraska	Pete Ricketts	70¢	Dan Osborn 29¢ · Cindy Burbank 0¢	Competitive
New Hampshire	Democratic party	89¢	Republican party 12¢	Competitive
New Jersey	Cory Booker	97¢	Justin Murphy 2¢	Safe
New Mexico	Ben Ray Luján	97¢	Larry E. Marker 2¢	Safe
North Carolina	Roy Cooper	91¢	Michael Whatley 8¢	Competitive
Ohio	Sherrod Brown	51¢	Jon Husted 48¢	Battleground
Oklahoma	Kevin Hern	97¢	Democratic party 2¢	Safe
Oregon	Jeff Merkley	98¢	David Brock Smith 2¢	Safe
Rhode Island	Democratic party	98¢	Republican party 2¢	Safe
South Carolina	Republican party	88¢	Annie Andrews 13¢	Watch list
South Dakota	Mike Rounds	94¢	Brian Bengs 6¢ · Julian Beaudion 1¢	Safe
Tennessee	Bill Hagerty	98¢	Marquita Bradshaw 4¢	Safe
Texas	James Talarico	53¢	Ken Paxton 48¢	Battleground
Virginia	Mark Warner	96¢	Bert Mizusawa 3¢	Safe
West Virginia	Shelley Moore Capito	98¢	Rachel Fetty Anderson 3¢	Safe
Wyoming	Republican party	98¢	Democratic party 2¢	Safe

Two notes on completeness. Louisiana's seat is up this cycle but Kalshi's per-state board does not carry it as a separate contested market, so it is absent above rather than omitted. And a handful of races carry a third candidate with a real price, most notably Nebraska, where an independent is the main challenger, and Montana, where an independent is quoted ahead of the Democratic candidate. Those are the two seats where a two-party frame would mislead you.

The Traps On This Board

Three places where the panel and the market are close enough that the disagreement is not worth much, and one where they split for a reason worth naming.

The reverse split at 1¢ is a curiosity both sides agree on. Every model priced it at 2% or less. Democrats winning the Senate while losing the House would need the national environment to move in opposite directions in the two chambers at once, which the modern record does not offer an example of.

Republicans holding both at 15¢, blend 13%. The panel is within a few points of the market, and the seats that sit highest here are the ones weighting redistricting most heavily. It is the honest tail, not an edge.

The Democratic sweep at 48¢, blend 28%. This is the panel's largest fade and the market's second-favorite outcome. The disagreement is not about the national mood, which both sides read the same way. It is about whether a four-seat parlay through Republican-leaning states clears at the rate the market implies.

The split at 39¢, blend 58%. Every seat on the panel ranked it first. The market ranks it second, by 9 cents.

Settlement Timeline

Election day	November 3, 2026
Contract settles	February 1, 2027, on which party controls each chamber
Next catalysts	Handicapper rating updates run continuously through the fall; the generic-ballot aggregators refresh daily; final district lines in the states still litigating their maps
Refresh policy	This board is re-scored when the story moves, and every verdict on it grades in public at settlement. Prices as of August 12, 2026.

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.

More on this: PrizePicks Power Play Vs Flex Play: Strategy Guide · England Vs Pakistan Test Odds: Is The Kalshi Draw Cheap? · Florida Governor Odds: Donalds Vs Jolly After The Primary · 2026 Governor Races: Every Kalshi Board We Track · GTA 6 PC Release Date: The Pattern Rockstar Never Breaks

The Bottom Line

The market and the panel are looking at the same country and reading the House identically. The whole argument is one chamber, and it comes down to a question that has an answer rather than a vibe: how many Republican-held Senate seats are actually winnable this cycle? The live ratings say roughly four, one of which is already leaning Democratic and three of which are coin flips in states that lean the other way. Democrats need all four, and they need to hold an open tossup in Michigan while doing it.

The market prices that parlay at 50 cents. Seven models, none of which saw that price, priced it between 25 and 32. One of those reads will be wrong, and in November it will be obvious which, on a scoreboard we keep in public either way.

Prices on this page are Kalshi's book. If you also trade on Polymarket, code OS4 gets new users a $50 trading bonus on a $10 deposit — affiliate link; terms as stated by Polymarket; 18+, availability varies by state.