A Monte Carlo simulation of all 35 Senate races on the 2026 ballot. For each race, we pool polling, fundamentals, fundraising, candidate strength, and our own expert call into a single projection, then run 20,000 simulated election nights to get win probabilities, projected margins, and a topline for the full chamber.
Republicans currently hold 53 seats; Democrats and independents hold 47. Of the 65 seats not up this cycle, 31 are held by Republicans and 34 by Democrats. Control is decided by the 35 seats on the ballot. A 50-50 Senate is broken by the sitting Republican vice president.
Ratings: Toss-up under 60% · Lean 60–75% · Likely 75–99% · Safe over 99%.
In plain terms: for each of the 11 genuinely competitive races, we treat polling, fundamentals, money, candidate strength, and our own judgment as separate, independently uncertain readings on the same race, then combine them into one projection weighted by how reliable each reading is. The 24 remaining seats get a single fundamentals-based projection since they aren't realistically in play. From there, we simulate the whole map 20,000 times, letting a national mood, a regional mood, and race-specific luck all move independently, to see how often each side ends up with the majority.
Polling average, structural fundamentals, fundraising and ad-reservation signal, candidate strength, and our own expert rating, each with its own uncertainty, pooled DerSimonian-Laird style with outlier dampening.
Every simulation draws a national environment shock, a regional shock, and race-specific noise, scaled by how elastic each race is to the national mood.
Each run produces a full map. Across all runs, we get each race's win probability and margin distribution, plus the topline seat count and control odds.