2026 Midterms Odds: The Senate Flipped in 30 Days
Published 18 September 2026 · Prices frozen 18 September 2026, 21:10 UTC
Key takeaways
- The Senate line moved 13 points in 30 days. Polymarket implied a 47.75% Democratic Senate on 19 August and 60.75% on 18 September. Republicans went from 51.00% to 39.00% over the same 125 price points.
- Two exchanges, four control lines, no gap worth trading. Polymarket sits between 1.00 and 1.50 points more Democratic than Kalshi’s midpoint on every line. Both books are liquid, which is why they agree.
- The seat ladder confirms the control price to within a third of a point. Polymarket’s eleven Republican Senate seat buckets normalise to 38.67% for 50 seats or more, against 39.00% in the Balance of Power book. Independent markets, the same answer.
- Democrats are likelier to win the votes than the chamber. The popular-vote ladder gives them 92.46% of winning the national House vote and the control book gives them 90.50% of the House. The 2-point wedge is the map.
- The market’s central case is a near-2018 wave. Median implied margin is D+8.4 on Polymarket and D+8.0 on Kalshi, just short of the D+8.6 Democrats posted in 2018.
Every 2026 midterms odds tracker on the first page of Google is either a live widget with nothing to say or a snapshot that has stopped being true. The one ranking first was last updated on 27 August and still shows Kalshi’s Senate market at 53% Republican to 47% Democratic — the wrong way round from where the money has been for four days. That matters, because $12.5m sits on Polymarket’s Balance of Power book and Kalshi holds over 13 million contracts of open interest on the House alone. We pulled every midterm control, seat-count and popular-vote market from both public APIs, derived the four chamber probabilities from Polymarket’s five-outcome book by hand, checked each against the independent seat ladders, and walked 30 days of CLOB history. Two things came out: the Senate moved 13 points in a month, and the two exchanges now agree so closely that the real disagreement has moved somewhere else.
Where the 2026 midterms odds stand tonight
Here is the whole board. Polymarket does not list “Democrats win the Senate” as a contract — it lists five mutually exclusive outcomes for the two chambers together, so the chamber numbers have to be added up. Kalshi does the opposite and lists each chamber separately. Those are the two columns below, and the gap column uses Kalshi’s bid-ask midpoint, which is the fairest single number to put against a Polymarket price.
| Outcome | Polymarket | Kalshi last | Kalshi bid/ask | Gap vs mid |
|---|---|---|---|---|
| Democrats win the House | 90.50¢ | 90.00¢ | 89 / 90¢ | +1.00 |
| Republicans hold the House | 9.25¢ | 11.00¢ | 10 / 11¢ | −1.25 |
| Democrats win the Senate | 60.75¢ | 59.00¢ | 59 / 60¢ | +1.25 |
| Republicans hold the Senate | 39.00¢ | 41.00¢ | 40 / 41¢ | −1.50 |
Four lines, four gaps, none larger than a point and a half. That is what two genuinely liquid books look like when both are paying attention — and it is a very different picture from the French presidential market, where Kalshi’s last trade was 71 days old. The midterms are the one election where Kalshi is not the junior book.
How to read a balance-of-power market
This is the step every tracker skips, and skipping it is how you end up publishing a House number that is obviously reversed. Polymarket’s Balance of Power book has five outcomes covering both chambers at once. To get a single-chamber probability you add the outcomes that contain it.
| Outcome | Price | Contains D House | Contains D Senate | 24h volume |
|---|---|---|---|---|
| Democrats Sweep | 60.00¢ | yes | yes | $164,530* |
| R Senate, D House | 30.50¢ | yes | no | — |
| Republicans Sweep | 8.50¢ | no | no | — |
| D Senate, R House | 0.75¢ | no | yes | — |
| Other | 0.25¢ | no | no | — |
So the Democratic House price is 60.00 + 30.50 = 90.50¢ and the Democratic Senate price is 60.00 + 0.75 = 60.75¢. The Republican sides are 9.25¢ and 39.00¢. The asterisked figure is the event-level 24-hour volume, which is how Polymarket reports this book. Note how small “D Senate, R House” is at 0.75¢: the market thinks a Democratic Senate without a Democratic House is close to impossible, which is the most useful thing this book tells you.
The Senate flipped in 30 days
This is the story, and it is invisible on any page that shows you today without showing you last month. We pulled 125 price points from the CLOB history endpoint covering 19 August to 18 September and derived the chamber lines at both ends.
| Line | 19 Aug | 18 Sep | Change | 30-day range |
|---|---|---|---|---|
| Democrats Sweep | 46.50¢ | 60.00¢ | +13.50 | 45.5 – 60.5¢ |
| R Senate, D House | 38.50¢ | 30.50¢ | −8.00 | 30.5 – 39.5¢ |
| Republicans Sweep | 12.50¢ | 8.50¢ | −4.00 | 8.5 – 14.5¢ |
| Democrats win the Senate | 47.75¢ | 60.75¢ | +13.00 | — |
| Republicans hold the Senate | 51.00¢ | 39.00¢ | −12.00 | — |
| Democrats win the House | 85.00¢ | 90.50¢ | +5.50 | — |
Thirty days ago the Senate was a Republican hold at 51.00¢ and today it is a Democratic pickup at 60.75¢. Both outcomes touched the extreme of their 30-day range on the final reading, so this is not a wobble that has already mean-reverted. Kalshi crossed to a Democratic favourite on 14 September, its first Democratic lead in that market since April, and the Silver Bulletin Deluxe forecast put Democrats at 59% on 18 September, calling it the highest of the cycle. Three independent estimates inside two points of each other is about as much corroboration as this kind of number ever gets.
The seat ladder agrees with the control book
Polymarket runs a completely separate market on the number of Republican Senate seats, with $2.93m of lifetime volume and its own traders. If the Balance of Power price is right, that ladder should imply the same thing. Republicans hold 53 seats now, so Democrats need a net gain of four to reach 51; a gain of three produces a 50–50 chamber that Republicans still control on the vice-president’s tie-break. The control line is therefore the probability that Republicans finish on 49 or fewer.
| Republican seats | Raw | Normalised | Cumulative | Implies |
|---|---|---|---|---|
| 47 or fewer | 33.50¢ | 32.88% | 32.88% | D control |
| 48 | 15.50¢ | 15.21% | 48.09% | D control |
| 49 | 13.50¢ | 13.25% | 61.33% | D control |
| 50 | 11.50¢ | 11.29% | 72.62% | R control (tie-break) |
| 51 | 10.00¢ | 9.81% | 82.43% | R control |
| 52 | 8.20¢ | 8.05% | 90.48% | R control |
| 53 | 4.10¢ | 4.02% | 94.50% | R control |
| 54 or more | 5.60¢ | 5.50% | 100.00% | R control |
The ladder puts Democratic control at 61.33% against 60.75% in the Balance of Power book, and Republican control at 38.67% against 39.00%. Two separate markets, different traders, different resolution sources, landing a third of a point apart. That is the strongest evidence you can get that the headline number is not one whale’s opinion. The 54-or-more row collapses the thin top of the ladder (54 at 2.65¢, 55 at 1.45¢, 56 at 0.95¢, 57+ at 0.55¢) so the table stays readable.
The House seat book is 4.55¢ short of itself
The same check on the House goes differently, and the reason is worth understanding before you quote any single bucket. The ten Republican House seat contracts sum to 95.45¢, not 100¢. That is a 4.55-point underround, the mirror image of a bookmaker’s margin, and it exists because this market has taken $300,364 in its entire life and $760 in the last day.
| Republican seats | Raw | Normalised | Control |
|---|---|---|---|
| Below 190 | 23.50¢ | 24.62% | D |
| 190–194 | 10.00¢ | 10.48% | D |
| 195–199 | 16.00¢ | 16.76% | D |
| 200–204 | 9.60¢ | 10.06% | D |
| 205–209 | 11.80¢ | 12.36% | D |
| 210–214 | 9.70¢ | 10.16% | D |
| 215–219 | 6.90¢ | 7.23% | straddles 218 |
| 220–224 | 3.50¢ | 3.67% | R |
| 225–229 | 2.40¢ | 2.51% | R |
| 230 or more | 2.05¢ | 2.15% | R |
Normalised, everything at 220 or above comes to 8.33%, against 9.25% for Republican House control in the Balance of Power book. The difference is the slice of the 215–219 bucket that lands on 218 or 219 — the market is implicitly putting about an eighth of that bucket on the controlling side. Coherent, not contradictory, but it only works because we normalised first: quote the raw 3.50¢ on “220–224” as a probability and you understate it by nearly 5%.
Winning the vote is not winning the House
Polymarket runs a ladder on the national House popular vote margin. Strip out its “Other” contract — 45.50¢ on a market with $0 of volume and $9 of resting liquidity, which is noise rather than a price — and the remaining thirteen buckets sum to 108.15¢. Normalise those and the probability Democrats win the national House vote is 92.46%.
The same exchange gives them 90.50% of actually winning the House. The gap is small but real, and it points the right way: a roughly 2-point band in which Democrats win more votes nationally and still do not win the chamber. That is the districting and distribution effect, priced. It is also a sanity check on the whole book — a market with Democrats likelier to win the House than the votes would be telling you something was broken.
| Democratic margin | Polymarket | Kalshi mid | Difference |
|---|---|---|---|
| At least 2% | 90.57% | 86.50% | +4.07 |
| At least 4% | 85.02% | 81.50% | +3.52 |
| At least 6% | 74.39% | 68.50% | +5.89 |
| At least 8% | 55.43% | 50.50% | +4.93 |
| At least 10% | 30.00% | 27.50% | +2.50 |
| At least 12% | 10.12% | 11.50% | −1.38 |
| At least 14% | 3.65% | 4.50% | −0.85 |
| At least 16% | 1.48% | 2.50% | −1.02 |
Interpolating for the median: Polymarket’s central case is D+8.4 and Kalshi’s is D+8.0. For scale, the 2018 midterms produced 53.4% to 44.8%, a margin of 8.6 points and the largest on record for a party out of power. The markets are pricing a wave a couple of tenths short of that one, and they are doing it from almost exactly where the likely-voter-adjusted generic ballot averages have sat all month, in a D+7.4 to D+8.2 band.
Where the two exchanges actually disagree
Read the difference column in that table again, because it is the only genuine disagreement in this entire comparison and it has a shape. Polymarket is 3.5 to 5.9 points more confident of a Democratic win in the 2-to-10-point range. Kalshi is 0.9 to 1.4 points more confident of a double-digit blowout and of the tail below D+2. The two curves cross between D+10% and D+12%.
In plain terms: Polymarket’s traders think a normal-sized Democratic wave is close to locked in and a landslide is unlikely. Kalshi’s quotes carry fatter tails in both directions — more room for a blowout, and more room for the whole thing to be much closer than anyone expects. That is a real difference of opinion about variance rather than about direction, which is why it survives even though the two exchanges agree to within 1.5 points on who wins.
One caveat, and it is a big one: Kalshi’s popular-vote book is quote-only. Seven of its eight contracts have never traded and the eighth has done 1,008. Those midpoints are a market maker’s opinion, not a crowd’s. Trust the control markets, where both books have millions behind them — and there the disagreement is nearly nil.
The volume comparison everyone gets wrong
Trackers routinely put Kalshi’s volume next to Polymarket’s and declare a winner. They are not the same unit. Polymarket reports volume in dollars of notional traded. Kalshi reports it in contracts. A contract that changes hands at 11¢ adds 1 to Kalshi’s count and $0.11 to Polymarket’s equivalent measure, so comparing the two raw numbers overstates Kalshi by roughly the inverse of the price.
| Book | Lifetime | 24 hours | Open interest | Unit |
|---|---|---|---|---|
| Polymarket Balance of Power | $12,545,272 | $164,530 | — | dollars |
| Kalshi House — Democratic | 11,720,884 | 191,293 | 7,666,991 | contracts |
| Kalshi House — Republican | 20,913,030 | 537,724 | 13,373,013 | contracts |
| Kalshi Senate — Democratic | 4,078,515 | 84,274 | 2,088,348 | contracts |
| Kalshi Senate — Republican | 5,087,333 | 50,762 | 2,908,215 | contracts |
Even after correcting for the unit, the honest conclusion is that Kalshi is the bigger venue on the midterms. Its Republican House contract alone carries 13.4 million contracts of open interest, and a position that size is not a market maker’s inventory. That is the reverse of what we found on international elections, and it is why the two books agree here and diverge there.
When will we actually know?
Kalshi lists a market on how long the count takes, which nothing else in this comparison covers, and it is the most practical thing on this page if you plan to stay up. Prices are the probability that the Associated Press has called the majority by each time.
| Called by | House | Senate | Basis |
|---|---|---|---|
| 11pm, 3 Nov | 12¢ | 12¢ | traded |
| Midnight, 4 Nov | 33¢ | 37¢ | traded |
| 1am, 4 Nov | 46¢ | 51¢ | House traded / Senate quote |
| 4am, 4 Nov | 75¢ | 71¢ | House traded / Senate quote |
| 6 Nov / 5 Nov | 82¢ | 87¢ | traded |
| 11 Nov / 14 Nov | 87¢ | 90¢ | traded |
| Into December | — | 95¢ | quote |
The House resolves fast: 46¢ by 1am Eastern, 75¢ by 4am and 82¢ by 6 November. The Senate is slower despite having far fewer races, and the tail is the interesting part — Kalshi lists contracts running to 5 January 2027 and still quotes roughly 5% that the majority is unresolved into December. That is runoff risk: Louisiana and Georgia can both push a close race past election night, and in a chamber that turns on four seats one runoff decides everything.
What to watch before 3 November
Watch the Senate seat ladder rather than the control price. Control is a step function on a single threshold and jumps for reasons you cannot see; the ladder shows the whole distribution moving and tells you whether a shift is broad or one race. Thirty-five seats are up this cycle, including special elections in Florida and Ohio, and 23 of them are Republican-held — that asymmetry is why a modest national swing moves this chamber so far.
Four races carry the Democratic path. North Carolina is the market’s most likely flip, Maine remains the cleanest toss-up on the board, the Ohio special is the highest-variance seat, and Iowa and Texas are where a genuine wave would first show up beyond the obvious targets. If three of those four hold as Democratic favourites through October, the 60.75¢ on this page will look low in hindsight.
Related on GeoOdds
- Balance of Power: 2026 midterms — live odds on the $12.5m control market
- Republican Senate seats after the midterms — live seat ladder
- House popular vote margin — live margin-of-victory odds
- North Carolina Senate 2026 — the market's most likely flip
- Maine Senate 2026 — the cleanest toss-up on the board
- Ohio Senate special 2026 — the highest-variance seat
- Polymarket vs Kalshi: French election odds — the same method where only one book still trades
New to prediction markets?
New traders get a $20 bonus
Deposit $10 on Polymarket and get a $20 trading bonus to put behind your first call.
See the midterm markets on Polymarket →New Polymarket users only. Bonus is trading credit, expires 30 days after it lands, and terms are set by Polymarket — check the current offer at signup. 18+. Fee-free markets. GeoOdds earns a commission on signups through this link.
Frequently asked questions
What are the 2026 midterms odds right now?
On 18 September 2026 Polymarket's Balance of Power market implies a 90.50% chance Democrats win the House and a 60.75% chance they win the Senate. Kalshi's two separate control markets last traded at 90¢ for the Democratic House and 59¢ for the Democratic Senate. The two exchanges are within 1.5 percentage points of each other on every one of the four control lines.
Are Democrats favoured to win the Senate in 2026?
Yes, and only recently. Polymarket's implied Democratic Senate probability was 47.75% on 19 August 2026 and 60.75% on 18 September, a 13-point swing in 30 days. Kalshi crossed to a Democratic favourite on 14 September, its first Democratic lead in that market since April. The Silver Bulletin Deluxe forecast had Democrats at 59% on 18 September, the highest of the cycle.
How many seats do Democrats need to win the Senate in 2026?
Four. Republicans hold a 53–47 majority, so Democrats must defend everything they have and flip four Republican-held seats to reach 51. A net gain of three produces a 50–50 chamber, which stays under Republican control because the vice-president breaks ties. That is why the market's Senate line tracks the probability that Republicans finish on 49 seats or fewer, not 50.
How many seats do Democrats need to win the House in 2026?
Three net flips from the 2024 result of 220 Republicans to 215 Democrats, which would put them at 218 of 435. The chamber has drifted since — 218 Republicans, 214 Democrats, one independent and two vacancies as of 14 September 2026 — but 218 is the number that matters and the seat-count market is priced around it.
Do Polymarket and Kalshi disagree about the 2026 midterms?
Barely, on control. Polymarket is between 1.00 and 1.50 points more Democratic than Kalshi's midpoint on all four chamber lines. The real disagreement is in the shape of the House popular-vote distribution: Polymarket is about 5 points more confident of a Democratic win of 6 to 10 points, while Kalshi puts more weight on a double-digit blowout. The two curves cross between D+10% and D+12%.
What margin do prediction markets expect in the House popular vote?
Polymarket's bucket ladder implies a median Democratic margin of 8.4 points and Kalshi's directional quotes imply 8.0 points. Both land just short of 2018, when Democrats won the House popular vote 53.4% to 44.8% for a margin of 8.6 points. They also sit almost exactly on the likely-voter-adjusted generic ballot averages, which have run between D+7.4 and D+8.2 this month.
Why do the House seat-count prices add up to less than 100?
Because the book is thin. Polymarket's ten Republican House seat buckets sum to 95.45¢, which is 4.55¢ underround — the opposite of a bookmaker's margin. That market has taken $300,364 in its lifetime and $760 in the last 24 hours, so the quotes are wide and nobody has bothered to arbitrage the gap. Treat the individual buckets as indicative and normalise them before you compare anything.
When will the 2026 midterm results be called?
Kalshi lists a market on it. Its House book last traded at 12¢ for an AP call before 11pm Eastern on election night, 46¢ by 1am, 75¢ by 4am and 82¢ by 6 November. The Senate is slower: 87¢ by 5 November and only 90¢ by 14 November, with contracts running out to 5 January. That long tail is the Louisiana and Georgia runoff risk, which can leave a close Senate unresolved for weeks.
Sources: Polymarket Gamma and CLOB public APIs (events 32228, 106193, 103166, 215753, 101856, 131798); the Kalshi public markets API (events CONTROLS-2026, CONTROLH-2026, KXHOUSEPOPVOTEMARGIND-26NOV03, KXAPCALLHOUSE-26AUG20, KXAPCALLSENATE-26AUG20); the Silver Bulletin 2026 forecast for the model comparison and the generic-ballot readings; Wikipedia on the 2018 House popular vote, and its 2026 Senate elections page on the seats up this cycle and the current 53–47 chamber. Prices captured 18 September 2026, 21:10 UTC and not updated since. Nothing here is investment advice — see our disclaimer.