Races · Senate · 2026 · AK

Senate · class II · open seat

Mary Peltola vs Dan Sullivan

AL Senate →

Lean D D +4.0 · 176 days to election · 9 polls · 13 markets Last poll 21d ago Markets 4d ago Rating 1d ago

Where this race stands

Verified

Lean D · model 89% R

lean-d · uncertainty-high

Forecast · poliagg-v8

Predicted final margin

R +17.0

80% CI: R +34.5 → D +0.5 · win prob 11%

R+30R+15TIEDD+15D+30

Quality · inputs

Data quality · medium

Polls used	9
Days to election	181
Residual σ	13.66pt
Generated	5/6/2026

▼ Compare with other models3 models

qgbt-v1GBT (qgbt)

D +0.8

80% CI R +6.0 → D +2.2

CV MAE 7.21

bayes-v1Bayes / Kalman

D +2.5

80% CI D +1.1 → D +3.9

CV MAE 1.10

poliagg-v8Ensemble (production)

R +17.0

80% CI R +34.5 → D +0.5

CV MAE 13.66

Polling average

Simple Weighted House-corrected Trimmed LOESS Kalman

Method:Weighted, plus each pollster's measured bias subtracted before averaging.

Show poll dots95% CI bandOverlay all methods|Toggle:

All polls · 9 results

Pollster:Population:Lean:High-quality only (weight ≥ 1.5)Sort:9 of 9 polls

End	Pollster	Weight	Lean	Sample	MoE	Pop	Trust signals	Results	Source
4/19/2026	Alaska Survey Research	1.00	—	1946	±2.5	LV	no scored pollsuncorrelated with editorial consensus+3 no scored polls Pollster has no historical polls with a known outcome in our database, so accuracy and bias cannot be measured. uncorrelated with editorial consensus Across 11 historical paired polls, this pollster's margins don't track the editorial consensus (Pearson r < 0.5). Independence from consensus is informational — could indicate genuine methodological independence or noisy results. ✓ verified 5d ago auto-checked on 5/4/2026. Deterministic checks confirmed bounds + structure (percentages in [0,100], totals plausibly close to 100%, dates ordered). Source page itself was not re-fetched. likely-voter screen Likely-voter screen — pollster filters respondents by reported turnout intent. Generally tighter for predicting elections than registered- or adult-population polls. n=1,946 Sample size of 1,946 respondents implies a margin of error around ±2.2pt — tighter than the ~600-respondent typical.	Dan Sullivan 43.0 · Mary Peltola 49.0	poll arch
3/22/2026	Alaska Survey Research	1.00	—	1590	±2.7	LV	no scored pollsuncorrelated with editorial consensus+3 no scored polls Pollster has no historical polls with a known outcome in our database, so accuracy and bias cannot be measured. uncorrelated with editorial consensus Across 11 historical paired polls, this pollster's margins don't track the editorial consensus (Pearson r < 0.5). Independence from consensus is informational — could indicate genuine methodological independence or noisy results. ✓ verified 5d ago auto-checked on 5/4/2026. Deterministic checks confirmed bounds + structure (percentages in [0,100], totals plausibly close to 100%, dates ordered). Source page itself was not re-fetched. likely-voter screen Likely-voter screen — pollster filters respondents by reported turnout intent. Generally tighter for predicting elections than registered- or adult-population polls. n=1,590 Sample size of 1,590 respondents implies a margin of error around ±2.5pt — tighter than the ~600-respondent typical.	Dan Sullivan 44.0 · Mary Peltola 49.0	poll arch
1/17/2026	Public Policy Polling	1.00	—(D+1.1)	611	—	unknown	commissioned by D-aligned: Internal D-aligned113d old+3 commissioned by D-aligned: Internal D-aligned This poll was commissioned by a Democratic-aligned committee, party, or PAC (Internal D-aligned). Such polls historically lean toward their sponsor. 113d old Poll was fielded 113 days ago. Older polls reflect an earlier state of the race and are weighted down in the rolling average. uncorrelated with editorial consensus Across 368 historical paired polls, this pollster's margins don't track the editorial consensus (Pearson r < 0.5). Independence from consensus is informational — could indicate genuine methodological independence or noisy results. ✓ verified 5d ago auto-checked on 5/4/2026. Deterministic checks confirmed bounds + structure (percentages in [0,100], totals plausibly close to 100%, dates ordered). Source page itself was not re-fetched. bias D+1.1pt Across 72 historical scored polls the average error vs. actual margin is D+1.1pt.	Dan Sullivan 47.0 · Mary Peltola 49.0	poll arch
1/11/2026	Alaska Survey Research	1.00	—	1988	±2.2	LV	no scored polls119d old+4 no scored polls Pollster has no historical polls with a known outcome in our database, so accuracy and bias cannot be measured. 119d old Poll was fielded 119 days ago. Older polls reflect an earlier state of the race and are weighted down in the rolling average. uncorrelated with editorial consensus Across 11 historical paired polls, this pollster's margins don't track the editorial consensus (Pearson r < 0.5). Independence from consensus is informational — could indicate genuine methodological independence or noisy results. ✓ verified 5d ago auto-checked on 5/4/2026. Deterministic checks confirmed bounds + structure (percentages in [0,100], totals plausibly close to 100%, dates ordered). Source page itself was not re-fetched. likely-voter screen Likely-voter screen — pollster filters respondents by reported turnout intent. Generally tighter for predicting elections than registered- or adult-population polls. n=1,988 Sample size of 1,988 respondents implies a margin of error around ±2.2pt — tighter than the ~600-respondent typical.	Dan Sullivan 46.0 · Mary Peltola 48.0	poll arch
10/23/2025	Data for Progress	1.00	—(D+1.0)	823	±3.0	LV	commissioned by D-aligned: Progress (D)199d old+4 commissioned by D-aligned: Progress (D) This poll was commissioned by a Democratic-aligned committee, party, or PAC (Progress (D)). Such polls historically lean toward their sponsor. 199d old Poll was fielded 199 days ago. Older polls reflect an earlier state of the race and are weighted down in the rolling average. high variance vs editorial consensus Stdev of (poll − rater consensus) across 127 historical paired polls exceeds 8pt — this pollster's results swing widely relative to where the editors place the same races. ✓ verified 5d ago auto-checked on 5/4/2026. Deterministic checks confirmed bounds + structure (percentages in [0,100], totals plausibly close to 100%, dates ordered). Source page itself was not re-fetched. likely-voter screen Likely-voter screen — pollster filters respondents by reported turnout intent. Generally tighter for predicting elections than registered- or adult-population polls. accurate · bias D+1.0pt Across 50 historical scored polls this pollster's average error vs. actual margin is D+1.0pt — within ±1pt.	Dan Sullivan 45.0 · Mary Peltola 46.0	poll arch
10/15/2025	Alaska Survey Research	1.00	—	1708	—	LV	no scored polls207d old+4 no scored polls Pollster has no historical polls with a known outcome in our database, so accuracy and bias cannot be measured. 207d old Poll was fielded 207 days ago. Older polls reflect an earlier state of the race and are weighted down in the rolling average. uncorrelated with editorial consensus Across 11 historical paired polls, this pollster's margins don't track the editorial consensus (Pearson r < 0.5). Independence from consensus is informational — could indicate genuine methodological independence or noisy results. ✓ verified 5d ago auto-checked on 5/4/2026. Deterministic checks confirmed bounds + structure (percentages in [0,100], totals plausibly close to 100%, dates ordered). Source page itself was not re-fetched. likely-voter screen Likely-voter screen — pollster filters respondents by reported turnout intent. Generally tighter for predicting elections than registered- or adult-population polls. n=1,708 Sample size of 1,708 respondents implies a margin of error around ±2.4pt — tighter than the ~600-respondent typical.	Dan Sullivan 46.0 · Mary Peltola 48.0	poll arch
8/1/2025	Alaska Survey Research	1.00	—	1623	±2.5	LV	no scored polls282d old+4 no scored polls Pollster has no historical polls with a known outcome in our database, so accuracy and bias cannot be measured. 282d old Poll was fielded 282 days ago. Older polls reflect an earlier state of the race and are weighted down in the rolling average. uncorrelated with editorial consensus Across 11 historical paired polls, this pollster's margins don't track the editorial consensus (Pearson r < 0.5). Independence from consensus is informational — could indicate genuine methodological independence or noisy results. ✓ verified 5d ago auto-checked on 5/4/2026. Deterministic checks confirmed bounds + structure (percentages in [0,100], totals plausibly close to 100%, dates ordered). Source page itself was not re-fetched. likely-voter screen Likely-voter screen — pollster filters respondents by reported turnout intent. Generally tighter for predicting elections than registered- or adult-population polls. n=1,623 Sample size of 1,623 respondents implies a margin of error around ±2.4pt — tighter than the ~600-respondent typical.	Dan Sullivan 47.0 · Mary Peltola 42.0	poll arch
7/27/2025	Data for Progress	1.00	—(D+1.0)	678	±4.0	LV	commissioned by D-aligned: Progress (D)287d old+4 commissioned by D-aligned: Progress (D) This poll was commissioned by a Democratic-aligned committee, party, or PAC (Progress (D)). Such polls historically lean toward their sponsor. 287d old Poll was fielded 287 days ago. Older polls reflect an earlier state of the race and are weighted down in the rolling average. high variance vs editorial consensus Stdev of (poll − rater consensus) across 127 historical paired polls exceeds 8pt — this pollster's results swing widely relative to where the editors place the same races. ✓ verified 5d ago auto-checked on 5/4/2026. Deterministic checks confirmed bounds + structure (percentages in [0,100], totals plausibly close to 100%, dates ordered). Source page itself was not re-fetched. likely-voter screen Likely-voter screen — pollster filters respondents by reported turnout intent. Generally tighter for predicting elections than registered- or adult-population polls. accurate · bias D+1.0pt Across 50 historical scored polls this pollster's average error vs. actual margin is D+1.0pt — within ±1pt.	Dan Sullivan 46.0 · Mary Peltola 45.0	poll arch
4/25/2023	Alaska Survey Research	1.00	—	1261	—	LV	no scored polls1111d old+3 no scored polls Pollster has no historical polls with a known outcome in our database, so accuracy and bias cannot be measured. 1111d old Poll was fielded 1111 days ago. Older polls reflect an earlier state of the race and are weighted down in the rolling average. uncorrelated with editorial consensus Across 11 historical paired polls, this pollster's margins don't track the editorial consensus (Pearson r < 0.5). Independence from consensus is informational — could indicate genuine methodological independence or noisy results. ✓ verified 5d ago auto-checked on 5/4/2026. Deterministic checks confirmed bounds + structure (percentages in [0,100], totals plausibly close to 100%, dates ordered). Source page itself was not re-fetched. likely-voter screen Likely-voter screen — pollster filters respondents by reported turnout intent. Generally tighter for predicting elections than registered- or adult-population polls.	Dan Sullivan 41.0 · Mary Peltola 44.0	poll arch

Trust signals are deterministic flags computed from the pollster scorecard (track-record size, historical bias, aggregation weight) and the poll itself (sample size, population, age, citation availability). No interpretation; click any row to see the cited numbers behind each flag.

Editorial ratings · poll-vs-rater

3 raters

Rater	Rating	Updated	Poll D-marg.	Rater marg.	Δ poll−rater
The Cook Political Report	Lean R	Jan 12	+4.0	-3.5	+7.5
Inside Elections	Lean R	Jan 12	+4.0	-3.5	+7.5
Sabato's Crystal Ball	Lean R	Jan 29	+4.0	-3.5	+7.5

Δ POLL−RATER = polling-consensus margin minus the rater's implied bucket midpoint. Positive = polls more D than rater; negative = rater more D than polls. |Δ| < 0.6 well-aligned · 0.6–1.5 mild divergence · ≥1.5 contested call.

last poll 3 weeks ago (4/19/2026) last market quote 4 days ago rating computed 1 day ago

Rating history (1 change)