Methodology

Data status

Forecasts and market probabilities

Research draft — not our official November call yet. Model hier-loeo-v1-candidate.

Projected Senate seats are the median seat count across model simulations: half the simulated outcomes fall at or below that count, and half at or above it. They are not called results or a range of possible outcomes. The historical two-party chamber series stores Democratic seats and probability; Republican values are their chamber-total and probability complements. Both parties appear equally in the charts. A 50–50 Senate split gives Republicans control through the Vice President.

Kalshi control percentages are market-implied probabilities. The Republican percentage is the complement of the Democratic quote.

Race-level Kalshi histories show separately priced Democratic and Republican Yes contracts, which may not sum to 100%. When one race quote is unavailable, the evidence card uses the available quote's complement.

The public site does not republish RealClearPolitics poll compilations or averages. Race cards, chamber maps, and desk overlays show a Phoenix Dataworks polling average: a recency-weighted two-party mean of desk-entered public polls in the last 21 days, with credit `Phoenix Dataworks average of N public polls (pollsters: …)`. Individual pollster toplines may still appear separately when we have licensed or desk-entered data, with credit to the original pollster. Arrow keys inspect dates on market charts and any public poll series we show.

The forecast model still uses RealClearPolitics polling averages and expert ratings as inputs. Publishing the combined probability is not the same as republishing those sources. Public `/api/control/forecast` omits the per-source `components` and `weights` breakdowns.