How we rate races
Rating tiers
Safe
>12 pts
Outcome not in doubt under any reasonable scenario.
Likely
6–12 pts
Strongly favors one party; requires a major shift to flip.
Lean
2–6 pts
Favored party has a clear but not decisive edge.
Tilt
0–2 pts
Essentially a toss-up with a directional lean.
Input signals
Modeled result
Primary signal. Structural baseline + polling + incumbency correction + national environment.
Polling average
TPSI weighted model (recency, sample size, pollster grade). Used when model result is unavailable.
Approval rating
Net approval of the incumbent in state. Informs open-seat baseline and candidate quality.
National environment
Generic ballot, presidential approval, and historical midterm wave patterns as a correction.
Correction factors
−8.5 pts
Generic ballot correction
Applied to Dem baseline to account for structural partisan lean per state.
+16.7 pts
Structural R adjustment
Republican structural advantage derived from 2024 cycle modeling.
Partial
Senate / Governor independence
Popular incumbents can significantly outperform their state's presidential baseline.
Live
Update cadence
Ratings update automatically as new polls enter the TPSI database.
Full rating scale
Safe D>12 pts D
Likely D6–12 pts D
Lean D2–6 pts D
Tilt D0–2 pts D
Toss-up±0 / no signal
Tilt R0–2 pts R
Lean R2–6 pts R
Likely R6–12 pts R
Safe R>12 pts R
Note: Toss-up is reserved for races where polling and structural signals are in direct conflict or unavailable. A "Tilt" has a measurable lean but remains within the margin of error. Modeled result convention: positive = R advantage, negative = D advantage.