2026 Midterm Forecast

U.S. Senate Forecast

Model-driven seat projections for all 35 Senate races. Updated continuously using polling, ratings, incumbency, and State Voting History

Model-driven - 1,000,000 pre-computed simulations using polling, ratings, national environment, and race volatility
How it works ↓
Federal Projection
Senate Majority Probability 35 seats on the ballot · 51 needed for majority  —
Democrat
—
majority chance
Calculating…
Republican
—
majority chance
Democratic Seats
—
Republican Seats
—
2026 Senate Map Click a state for details

Loading Interactive Map…

Color
D (60–100%)
Toss-up (very light)
R (60–100%)
No Race
Projected Flip
Senate Tipping Point The Tipping Point is the specific "win" in the middle of that snake that pushes a party over the 50-seat line to claim the majority
Projected
—·— 51 = majority
D (60%→Safe) Toss-up (very light) R (60%→Safe) Tipping point
All 2026 Races
StateRatingMarginWin ProbIncumbent
How the Model Works
Four-Stage Pipeline
1

Collect Signals

Gather expert ratings, polling averages, the state voting history, and incumbency status for all 35 races.

2

Composite Margin

Normalize signal weights to a true weighted average, then add an incumbency bonus that scales with how competitive the race is.

3

Win Probability

Convert the margin to a win probability using a normal distribution with an 8.5-point uncertainty margin.

4

Senate Control

Run 1,000,000 correlated Monte Carlo simulations to determine how often each party hits 51 seats.

Signal Weighting
Expert Ratings (40%)
Polling Averages (35%)
State Voting History (15%)
Incumbency Bonus (10%)

Note: The three base signals (ratings, polling, history) are normalized to a true weighted average. The incumbency bonus is then added on top — scaled back only when live polling shows a wide margin, and applied in full when no polling is available. For open seats, the incumbency bonus is removed entirely and its weight is spread across the remaining signals.

1,000,000 Pre-Computed Simulations: Senate control probabilities are computed offline using 1 million correlated Monte Carlo simulations with a fixed seed for consistency — the same input data always produces the same result. Each simulation applies a shared national environment shift (σ = 8.5 pts) and independent race noise (σ = 8.5 pts). Results are model projections, not guarantees.