2026 Senate Odds: Who Wins the Midterms, According to the Markets

What are the odds of the Democrats winning the Senate in 2026?

As of August 24, 2026, Polymarket gives Democrats a 50.5% chance of controlling the Senate and Republicans 49.5%. A coin flip. The House is a different animal: Democrats are 88.5% there.

That Senate number is new. When I published this on August 10, Republicans were favored at about 53 cents, and I put 20 units on "R Senate / D House" at 40 cents. Two weeks later the control market has flipped a hair blue. My position is marked down to 37.5, and I'm not touching it, for reasons the rest of this post explains. One mechanical detail everyone forgets: a 50-50 Senate stays Republican because Vice President Vance breaks ties, so Democrats effectively need 51 seats, not 50.

The thing that looks broken

Open Polymarket's Balance of Power market right now and you'll see two things that seem like they can't both be true.

When I placed the bet, the control market priced "Republicans keep the Senate" around 53 cents. Fine. Then you clicked into the ten closest individual Senate races, and Democrats were favored in eight of them. Now it's nine of ten (Iowa is the holdout). If you just picked the favorite in every race (bettors call this "chalk"), Democrats win the Senate.

When I first saw that I thought: free money, somebody's wrong. I said as much on Episode 9. So I did what I always do when a market looks broken (and what I should have done before some other bets we don't need to relive): I actually did the math.

The race board

Prices are Polymarket midpoints as of August 24, 2026. Ratings from Cook Political Report (Aug 5). Poll column last refreshed August 10. This table gets refreshed as the race moves.

Race	D win %	Cook rating	Freshest poll (30 days)
Iowa: Turek v Hinson	40.5	Lean R	Hinson +3 (Emerson, Aug 6)
Texas: Talarico v Paxton	50.5	Lean R	Talarico +4 (TX Pulse, Aug 7)
Ohio special: Brown v Husted	53.5	Toss Up	none
Alaska: Peltola v Sullivan	57.0	Toss Up	none
Michigan: El-Sayed v Rogers	64.0	Toss Up	tie 45-45 (Mitchell, pre-primary)
Maine: Jackson v Collins	69.5	Toss Up	Jackson +3 (UNH)
New Hampshire: Pappas v R field	85.5	Lean D	none
Minnesota: Flanagan v R field	90.5	Likely D	Flanagan won D primary Aug 11
North Carolina: Cooper v Whatley	92.5	Lean D	Cooper +9 (Fox)
Georgia: Ossoff v Collins	93.0	Lean D	Ossoff +9 (AARP)

Notice something? Nine of the ten lean Democrat now, sure. But look at how much. Texas is a literal coin flip. Ohio is 53.5. Alaska is 57. "Favored" at 53 cents isn't favored in any way your gut understands. That's the whole trick of this market. (Also note the "none" entries: some of these races have close to zero recent public polling. Those prices are trader vibes.)

Add up every race and the paradox mostly disappears

Here's the exercise nobody does. Take every seat, weight it by the market's probability, and count the expected seats.

Democrats hold 47 now. Their safe seats stay put. Sum the ten competitive races above and the markets expect Democrats to win about 7.0 of them. Run the full ledger and you get an expected Senate of roughly:

So Democrats don't just need to win the nine races they lead. They need those nine plus nothing going wrong anywhere else, and several of those "leads" are 50-57 cent leads. Price that honestly and a control market sitting at even money isn't a contradiction of the race markets. It's approximately what the race markets already say. The extra half-cent Democrats now carry in the control market is traders paying a small premium for the scenario where the close races all break the same way.

The markets aren't broken. What's actually being priced is a real question: are these races independent coin flips, or do they all swing together?

The correlation bet (this is the actual bet)

Senate races aren't independent. If Democrats win Texas, it's probably because there's a wave, and if there's a wave they're probably also winning Ohio and Alaska. It's like Patrick Mahomes throwing four touchdowns: you don't really need to check whether the Chiefs won.

Waves are how you get a Democratic sweep, which is why "Dem sweep" trades at 46.5 cents, the priciest single outcome in the market. My 20 units on R Senate / D House at 40 cents is, mechanically, a bet against perfect correlation: blue enough of a year that the House flips (the market has that at 88.5%, and I agree it's close to a lock), but not the once-a-generation wave that also drags Texas, Ohio, Alaska AND Maine across the line while Democrats hold every seat of their own.

Why I don't believe in the full wave: redistricting

The generic ballot said Democrats +7 when I placed the bet, and sits around +6 as I update this. That's still wave-adjacent territory, roughly the August 2018 number, and 2018 ended at D+41 House seats. But 2018 didn't happen on this map. Mid-decade redistricting added about a dozen Trump-won districts (Texas, Missouri, North Carolina, Ohio) against roughly five new Harris-won ones from California. I had my AI run the numbers and it reckons the new maps blunt a D+7 environment by 8 or 9 House seats. My guess is Democrats net around 20 seats, not 40. Enough to flip the House comfortably. Not the tsunami the sweep price needs.

And one more thing the wave crowd skips: in 2010, 2014 and 2022, the GOP beat its August generic-ballot number every time. 2018 is the only recent year the August lead held. I'm not paying 46 cents to bet on the exception.

My position, and what makes me trade out

20 units on R Senate / D House at 40 cents, entered around August 8. Marked at 37.5 as of this update, so I'm down about 6%, which also means nothing. The control market flipping blue hurts the mark, but my thesis was never "Republicans are safe." It was "the House flips and the Senate doesn't quite," and a dead-heat expected-seats tally is fine for that. I also still hold 10 units on Democrats House from Episode 2, bought at 83 cents, so yes, I'm doubled up on the House side. That's deliberate. The House is the strong leg; the Senate is where the price is.

The exit plan, since I keep learning this lesson in public: if this position runs into the high 50s before November, I'll probably sell half and let the rest ride to resolution. If Ohio and Alaska start polling D+5, or the generic ballot pushes past +10, that's the wave scenario, my thesis is wrong, and I'll cut it rather than argue with it.

The numbers will move. This post will too.

Ninety days is forever in an election. We'll update the race board and the expected-seats math as prices move; check the date stamp up top. If you want to translate any of these cent-prices into odds you're used to, that's what our converter is for.

Hear the full midterms breakdown on Episode 9, including the Michigan primary, the redistricting fight, and why Andrew agrees with the split. We picked it up again on Episode 10: the House position is in at 81¢ and now trades at 89, and the Senate sits at 51/51.