The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is +1.0 points across 6 election cycles.

2024 presidential result

R +1.4

Outside money, two weeks

$29.8M D-side · $32.5M R-side

Extra money vs typical

+0.55 SD

Heat, D / R

3.93 (3) / 3.70 (2)

Range, about two in three chance

± 4.5 points: 1.9 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is +0.6 points across 6 election cycles.

2024 presidential result

R +2.2

Outside money, two weeks

$2.9M D-side · $0.4M R-side

Extra money vs typical

-0.91 SD

Heat, D / R

3.90 (13) / 3.85 (13)

Range, about two in three chance

± 4.5 points: 1.9 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is +0.5 points across 6 election cycles.

2024 presidential result

R +13.7

Outside money, two weeks

$40.6M D-side · $71.6M R-side

Extra money vs typical

+1.03 SD

Heat, D / R

3.75 (2) / 4.20 (2)

Range, about two in three chance

± 4.5 points: 1.9 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is +0.8 points across 5 election cycles.

2024 presidential result

R +3.2

Outside money, two weeks

$1.2M D-side · $1.6M R-side

Extra money vs typical

-0.22 SD

Heat, D / R

3.10 (2) / 3.50 (2)

Range, about two in three chance

± 4.5 points: 1.8 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is -0.3 points across 6 election cycles.

2024 presidential result

D +6.9

Outside money, two weeks

$15.3M D-side · $8.5M R-side

Extra money vs typical

-1.36 SD

Heat, D / R

3.70 (2) / 2.85 (2)

Range, about two in three chance

± 4.5 points: 1.9 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is -1.9 points across 6 election cycles.

2024 presidential result

R +11.2

Outside money, two weeks

$48.9M D-side · $40.7M R-side

Extra money vs typical

+1.55 SD

Heat, D / R

3.17 (3) / 3.85 (2)

Range, about two in three chance

± 4.5 points: 2.0 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is -1.5 points across 6 election cycles.

2024 presidential result

R +13.1

Outside money, two weeks

$14.8M D-side · $8.5M R-side

Extra money vs typical

-0.71 SD

Heat, D / R

2.68 (12) / 3.45 (11)

Range, about two in three chance

± 4.5 points: 1.9 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is -1.8 points across 6 election cycles.

2024 presidential result

R +13.2

Outside money, two weeks

$40.3M D-side · $6.6M R-side

Extra money vs typical

+0.09 SD

Heat, D / R

3.33 (4) / 2.05 (2)

Range, about two in three chance

± 4.5 points: 1.8 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is +0.3 points across 6 election cycles.

2024 presidential result

R +0.9

Campaign digital ads, ~4 weeks

$0.9M D-side · $2.4M R-side

Extra money vs typical

+1.44 SD

Heat, D / R

1.50 (2) / 2.40 (2)

Range, about two in three chance

± 4.5 points: 2.0 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is +1.0 points across 6 election cycles.

2024 presidential result

R +1.4

Campaign digital ads, ~4 weeks

$0.1M D-side · $0.3M R-side

Extra money vs typical

-0.19 SD

Heat, D / R

2.70 (2) / 3.90 (2)

Range, about two in three chance

± 4.5 points: 1.8 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is +0.6 points across 6 election cycles.

2024 presidential result

R +2.2

Campaign digital ads, ~4 weeks

$0.0M D-side · $4.2M R-side

Extra money vs typical

+0.43 SD

Heat, D / R

1.70 (2) / 1.60 (2)

Range, about two in three chance

± 4.5 points: 1.9 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is -1.9 points across 6 election cycles.

2024 presidential result

R +11.2

Campaign digital ads, ~4 weeks

$0.7M D-side · $2.5M R-side

Extra money vs typical

-1.21 SD

Heat, D / R

3.50 (2) / 3.80 (2)

Range, about two in three chance

± 4.5 points: 1.9 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.

The adjustment starts from the polling read above and adds two terms. Recent polling error: a lean toward Democrats, sized for each state. Spending: outside money in the two weeks to 2026-10-26, compared with what a race this close usually draws, read against the state’s 2024 presidential result. Rhetoric is scored but not counted: it has no track record yet, so it is frozen on November 2 and graded on its own.

Spending counts 1.55 points per standard deviation of extra money, a starting assumption of 1.5 updated with 43 Senate races from 2020 and 2022. Rhetoric counts 0.75 points per point of heat gap, a starting assumption with no history behind it, capped at 1.5 points. The national polling-error lean is 2.5 points, give or take 1.8: in every off-year since 1999, about a third of that year’s polling miss repeated in the next year’s elections, and the 2025 miss was 7.5 points toward Democrats. That history is thin, so the range above includes this uncertainty, along with the race-by-race error of a final polling average, about 4 points once the year’s shared miss is set aside. The true result should land inside the range about two times in three. Each state then moves by 0.69 times its own average miss against the national miss over the last 12 years; here that is -1.8 points across 6 election cycles.

2024 presidential result

R +13.2

Campaign digital ads, ~4 weeks

$1.8M D-side · $0.8M R-side

Extra money vs typical

-0.47 SD

Heat, D / R

2.60 (2) / 3.55 (2)

Range, about two in three chance

± 4.5 points: 1.9 from our adjustments, 4.1 from this race’s own polling error

Outside spending is FEC independent expenditures only; campaigns’ own ad budgets and reservations are not included. Heat scores hyperbole, which stands in for attack rhetoric. Experimental. Recent polling error and outside spending are read as signs of where campaigns' internal polls stand. It is a bet, frozen on November 2 and graded against certified results beside the polls. Rhetoric is scored and graded separately, not counted.