Choose lower or upper chambers. Each state's fill identifies the party holding the larger seat share; stronger majorities deepen the color. Select a state for its exact composition, district map, and closest races.
All 5,743 scheduled contests and 99 chamber resources are wired. The active edition publishes 78 numeric control distributions, with seven chambers explicitly withheld, three structural systems outside a binary D/R model, and eleven legal schedule carries.
Chambers whose projected controlling party differs from the party holding them today. A chamber whose call is unresolved is not counted, and neither is a chamber whose control forecast the release withholds — an absent forecast is not a projection of a change. Both the direction and the rating come from the model row itself, never from the hand-set consensus file that sits beside it. The rating says how confident the change is, which for a close chamber can still be a toss-up.
Control forecasts from the current modeled field.
The scale reads chance of chamber control — not a seat margin: R control at the left rail, D control at the right. Chambers past 99 pin at the rail; the printed number is the honest one.
Each rating is the model's own control probability, bucketed on the same scale every probability on this site uses — Solid at 95 in 100 and above, Likely from 75, Lean from 60. It is not a separate editorial call.
| New Hampshire Senate | Tossup | R+8 (8 D vs 16 R) | D 49% | 13 of 24 (80%: 9–15) |
| Michigan Senate | Tossup | D+2 (20 D vs 18 R) | R 46% | 19 of 38 (80%: 16–21) |
| Wisconsin Senate | Lean R | R+3 (15 D vs 18 R) | R 63% | 16 of 33 (80%: 15–18) |
| Minnesota House | Lean D | R+0 (67 D vs 67 R) | D 65% | 69 of 134 (80%: 66–77) |
| Pennsylvania House | Lean R | D+1 (102 D vs 101 R) | R 67% | 98 of 203 (80%: 87–104) |
| Arizona Senate | Tossup | R+4 (13 D vs 17 R) | R 55% | 14 of 30 (80%: 12–16) |
| Michigan House | Lean R | R+6 (52 D vs 58 R) | R 66% | 53 of 110 (80%: 44–55) |
| Minnesota Senate | Lean D | D+1 (34 D vs 33 R) | D 71% | 36 of 67 (80%: 32–39) |
| Wisconsin Assembly | Lean R | R+9 (45 D vs 54 R) | R 73% | 45 of 99 (80%: 36–54) |
| Georgia House | Likely R | R+20 (80 D vs 100 R) | R 75% | 86 of 180 (80%: 80–89) |
| Maine Senate | Likely D | D+5 (20 D vs 15 R) | D 79% | 20 of 35 (80%: 17–23) |
| Maine House | Likely D | D+3 (76 D vs 73 R) | D 79% | 84 of 151 (80%: 74–90) |
| Vermont Senate | Likely D | D+3 (16 D vs 13 R) | D 80% | 21 of 30 (80%: 25–29) |
| Arizona House | Lean R | R+6 (27 D vs 33 R) | R 73% | 27 of 60 (80%: 24–31) |
| North Carolina House | Likely R | R+22 (49 D vs 71 R) | R 82% | 53 of 120 (80%: 46–55) |
| North Carolina Senate | Likely R | R+10 (20 D vs 30 R) | R 83% | 21 of 50 (80%: 18–22) |
| Georgia Senate | Likely R | R+10 (23 D vs 33 R) | R 84% | 25 of 56 (80%: 23–25) |
| New Mexico House | Likely D | D+18 (44 D vs 26 R) | D 91% | 43 of 70 (80%: 40–48) |
| California — official-field simulation | ||||
These chambers have district resources and maps, but the active edition does not publish a numeric chamber-control distribution. No legacy probability is substituted.
No D/R forecast exists for these chambers — a two-party control probability would misdescribe how they are governed. What follows are facts, not a forecast.
These chambers hold no election this year, so control carries by legal schedule — a fact of the calendar, not a simulation.
Governor plus both legislative chambers, colored by projected control. Hover any state for its three components — each shown as a published control chance, a not-up-in-2026 fact, or an honest no-forecast label. Nebraska is shown separately as a Republican governor with an officially nonpartisan legislature—not as missing data.
Fill depth follows the weakest available component confidence in the governor-and-legislature stack. Not-up offices carry their current party as fact; bounded and nonpartisan systems remain explicitly labeled.
This is one of our least certain forecasts — state-legislative races have almost no district polling and many of these maps were redrawn. Read the range, not the midpoint. How this is modeled →