Balance of power watch
Thirty-five Senate seats, the whole House and thirty-six governorships. Which of them are on the ballot and who holds them now is public record. What happens to them is this desk's opinion — and every one of those opinions is published here with a number attached, months early, so the certified result can grade it.
The map does most of the work. Democrats must win every toss-up and then take a seat the desk currently rates as leaning the other way, without losing Georgia.
The midterm prior against the president's party is enormous and the majority is four seats. The counterweight is redistricting, which is why this is 63 and not 80.
Eighteen seats each and an unusual number of open mansions; the toss-ups tilt marginally toward the party out of the White House.
These three numbers are the whole exposure of the desk this cycle. They are dated, they will move as the inputs move, and every move is logged with its reason.
Ratings are opinion. Platforms are not. The Candidate Ledger sets the Georgia field side by side under identical headings, with a first-party source and a date on every line and nothing weighed against anything.
The chamber below is drawn at the desk's current ratings: holdovers plus every seat rated anything other than a toss-up. The pale block in the middle is the part of the Senate this election genuinely decides.
22 Republican-held and 13 Democratic-held seats are on the ballot, including two specials for appointed senators. With the vice-presidency breaking ties, fifty holds the chamber for Republicans and fifty-one is required to take it.
The single most exposed incumbent of the cycle: a first-termer who won by a fraction in a state that voted the other way in 2024.
The last Republican holding a seat in a state the other party carried nationally; every cycle has been called wrong here.
An open seat in the closest state in the country for three cycles running.
Iowa is redder than it was but Ernst's margins have never been comfortable, and midterms punish the president's party here.
Michigan's last two Senate races were decided inside two points, and this one has no incumbent.
An open seat in a state that elects Republican governors and Democratic senators without apparent difficulty.
An appointed senator, a state that has moved right, and a midterm wind pointing the other way.
Texas is not close on the fundamentals; a nominee chosen by a low-turnout primary is the live risk.
Ranked-choice voting keeps Alaska from being safe for anyone, but Sullivan starts with the seat and the machine.
An appointed senator facing voters for the first time, in a state that has moved steadily right since 2018.
Kansas elects Democratic governors and Republican senators; the split has been stable.
Open seats invite primaries, and a bruising primary is the only route by which this seat becomes live.
The risk to Cassidy is a primary, not a general.
Minnesota keeps being called competitive and keeps not being competitive, but an open seat narrows it.
Hyde-Smith has underrun the state's baseline before, which is the only reason this is not safe.
An independent challenge outperformed expectations in 2024; the desk is not treating that as noise.
Graham attracts enormous opposition money and wins anyway.
The president's party has lost House seats in 19 of the last 21 midterms. That is the single strongest prior in American electoral politics and it is not a partisan claim.
Mid-decade redistricting in several states means the map the country votes on in November is not the map it voted on in 2024. Until every map is final, a national seat forecast is a guess dressed as arithmetic.
Roughly 30 seats decide control. The other 405 are decided already, whatever the coverage suggests.
District-level ratings arrive when the maps are final. Publishing a seat-by-seat House forecast against boundaries that are still in litigation would be a number the reader cannot trace, and the Standard forbids it.
The ratings above are the desk's judgement. This is the arithmetic, and the two are kept apart on purpose. Certified 2024 state margins go in, four printed coefficients move them, a shared national swing and a per-seat error are drawn 20,000 times, and every number below falls out of that. The randomness is seeded, so the figure you are reading is the figure everyone else is reading.
Median outcome: 49 Democratic seats, with eight runs in ten landing between 46 and 51. Fifty–fifty is not a tie — the vice-president breaks it for the Republicans, so 51 is the number that matters and the chart is drawn against 51.
- Ohio13.6%
- Florida13.0%
- Alaska10.5%
- Iowa10.2%
- Texas10.1%
- Kansas6.5%
- North Carolina6.1%
- Michigan4.9%
| Seat | 2024 | Lean | 1σ | Model — left | Desk rating |
|---|---|---|---|---|---|
North Carolina R held · open | 3.2ʀ | 3.1ʟ | ±7.0 | 66% | 50%(+16) |
Michigan D held · open | 1.4ʀ | 4.5ʟ | ±7.0 | 71% | 65% |
Ohio R held · special | 11.2ʀ | 4.1ʀ | ±5.5 | 26% | 35% |
Florida R held · special | 13.1ʀ | 5.6ʀ | ±5.5 | 19% | 15% |
New Hampshire D held · open | 2.8ʟ | 7.8ʟ | ±7.0 | 84% | 65%(+19) |
Georgia D held | 2.2ʀ | 6.9ʟ | ±5.5 | 85% | 50%(+35) |
Minnesota D held · open | 4.2ʟ | 8.9ʟ | ±7.0 | 88% | 85% |
Iowa R held | 13.2ʀ | 7.7ʀ | ±5.5 | 12% | 35%(-23) |
Alaska R held | 13.1ʀ | 7.6ʀ | ±5.5 | 12% | 15% |
Maine R held | 6.9ʟ | 8.0ʟ | ±5.5 | 89% | 50%(+39) |
Texas R held | 13.7ʀ | 8.1ʀ | ±5.5 | 10% | 35%(-25) |
Kansas R held | 16.1ʀ | 10.0ʀ | ±5.5 | 6% | 15% |
South Carolina R held | 17.9ʀ | 11.3ʀ | ±5.5 | 4% | 15%(-11) |
Illinois D held · open | 10.9ʟ | 14.1ʟ | ±7.0 | 97% | 97% |
Montana R held | 19.9ʀ | 12.9ʀ | ±5.5 | 2% | 3% |
Virginia D held | 5.8ʟ | 13.1ʟ | ±5.5 | 98% | 97% |
New Jersey D held | 5.9ʟ | 13.2ʟ | ±5.5 | 98% | 97% |
New Mexico D held | 6.0ʟ | 13.3ʟ | ±5.5 | 98% | 97% |
Nebraska R held | 20.5ʀ | 13.4ʀ | ±5.5 | 2% | 15%(-13) |
Louisiana R held | 22.0ʀ | 14.6ʀ | ±5.5 | 1% | 15%(-14) |
Alabama R held · open | 30.5ʀ | 18.2ʀ | ±7.0 | 1% | 3% |
Mississippi R held | 22.9ʀ | 15.3ʀ | ±5.5 | 1% | 15%(-14) |
Kentucky R held · open | 30.5ʀ | 18.2ʀ | ±7.0 | 1% | 15%(-14) |
Colorado D held | 11.0ʟ | 17.2ʟ | ±5.5 | 99% | 97% |
Rhode Island D held | 13.8ʟ | 19.3ʟ | ±5.5 | 100% | 97% |
Oregon D held | 14.3ʟ | 19.8ʟ | ±5.5 | 100% | 97% |
Delaware D held | 14.7ʟ | 20.1ʟ | ±5.5 | 100% | 97% |
South Dakota R held | 29.2ʀ | 20.2ʀ | ±5.5 | 0% | 3% |
Tennessee R held | 29.7ʀ | 20.6ʀ | ±5.5 | 0% | 3% |
Arkansas R held | 30.6ʀ | 21.3ʀ | ±5.5 | 0% | 3% |
Oklahoma R held | 34.3ʀ | 24.1ʀ | ±5.5 | 0% | 3% |
Idaho R held | 36.5ʀ | 25.9ʀ | ±5.5 | 0% | 3% |
West Virginia R held | 41.9ʀ | 30.1ʀ | ±5.5 | 0% | 3% |
Wyoming R held | 45.8ʀ | 33.1ʀ | ±5.5 | 0% | 3% |
Massachusetts D held | 25.2ʟ | 28.3ʟ | ±5.5 | 100% | 97% |
Margins are written as magnitude and side — 6.0ʟ is six points to the left column, 6.0ʀ six to the right. “Model — left” is the share of 20,000 simulations in which the left column wins the seat.
The written ratings were published before this model existed. Where the two differ by ten points or more, both numbers stay on the page and both get graded separately. Editing one to match the other after the fact is the move the Standard exists to prevent.
- Mainemodel 89% · desk 50%
- Georgiamodel 85% · desk 50%
- Texasmodel 10% · desk 35%
- Iowamodel 12% · desk 35%
- New Hampshiremodel 84% · desk 65%
- North Carolinamodel 66% · desk 50%
- Kentuckymodel 1% · desk 15%
- Mississippimodel 1% · desk 15%
- Louisianamodel 1% · desk 15%
- Nebraskamodel 2% · desk 15%
- South Carolinamodel 4% · desk 15%
Senate results track the presidential baseline of their state closely but not perfectly — statewide candidates still run a few points off the top of the ticket. The state's 2024 certified margin is carried at 0.78 rather than 1.0 so the model does not assume a Senate race is a presidential rerun.
In the 19 midterms since 1946 the president's party has lost national vote share almost every time. 5.6 points is the central tendency of that series, applied uniformly rather than tuned per state, because a uniform prior is the one a reader can check.
Name recognition, casework and money. Applied only where a sitting senator is actually running; retirements get nothing, which is why open seats move toward the state's baseline.
A senator who was appointed rather than elected carries a thinner version of incumbency. The bonus is reduced, not removed — they still hold the seat and the machine.
One draw per simulation, shared by all 35 seats. This is the correlation that makes a bad night a bad night everywhere at once — a model with independent seats produces a confidently wrong, far too narrow distribution.
Candidate quality, primaries, scandals, local drift. Independent per seat.
No incumbent means no track record, and an unresolved primary means the candidate is not yet a known object.
- No polling is in the model. When these races have a public polling record, polling enters here as its own coefficient, with its own line in the table, on a dated version bump.
- The midterm drag is a single national number, not a per-state estimate. States where the president's party has historically bled faster than average are not being credited for it.
- Candidate quality is inside the per-seat error term rather than being estimated. A badly chosen nominee shows up as variance, not as a moved central estimate.
- The 2024 presidential baseline is two years old and cannot see any drift since. That staleness is one of the things the state error term is paying for.
An unusually even cycle: 18 Republican-held and 18 Democratic-held mansions, with a large number of them open because the incumbent is termed out. Only the live ones are rated.
An open mansion in the closest state in the country, with no incumbent advantage on either side.
Open, three-way risk, and a state that has swung hard in both directions since 2016.
A Republican governor in a state whose electorate is still narrowly the other way.
Wisconsin decides statewide races by a point; removing the incumbent removes the only thing that was not a coin flip.
Hobbs won by a fraction of a point and has governed against a legislature of the other party ever since.
Open, in a state where the president's party has historically bled in midterms.
Kansas elects Democratic governors when Republicans nominate badly. That is a bet on the other side's mistake, not on the state.
Open, and Maine's ranked-choice general elections rarely resolve the way the first-round polling suggests.
An open mansion, but Florida has not elected a Democrat statewide since 2018.
New Hampshire re-elects its governors almost as a habit.
Ohio's statewide baseline has moved far enough right that an open seat is not the same as a competitive one.
One of the highest-approval governors in the country seeking a second term.
Every band on this board is a probability, not a mood. "Lean" without a number cannot be wrong, which is exactly why it is the industry standard. Ours carries a figure that the result will score.
- A generic-ballot margin outside four points that holds for a month, rather than a single survey with a friendly sample.
- Final maps in the mid-decade redistricting states landing more than five net seats away from the current draw.
- Special-election overperformance running consistently above the historical relationship with the eventual national margin — the earliest honest signal there is.
- A retirement in any seat currently rated Safe. Incumbents leave when they can count.
- Presidential approval crossing 45 or falling under 38 and staying there through Labor Day.
Ratings published early and dated
Every seat carries a probability from the day it appears on this board. Changes are logged with the date and the reason, never edited silently.
The board grades itself
On 4 November every rating on this page is scored against the certified result and folded into the desk's permanent calibration record — including the ones that embarrass us.
Structure and opinion are never mixed
Which seats are up and who holds them is public record. What happens to them is the desk's opinion, labelled as such wherever it appears.
Certified returns only, on the night
No projections sourced from other outlets' decision desks. When the canvass is not in, the board says the canvass is not in.
Structure: sitting members, class-2 Senate seats, 2026 gubernatorial cycle · retrieved 2026-08-04
Seat counts, incumbents and announced retirements are public record as of the retrieval date. Ratings are the desk's own and are labelled as ratings everywhere they appear.