ElectionTracker Intelligence.
An inspectable polling research project, built from published surveys rather than editorial guesses. See each poll, compare approaches, and understand what the numbers can and cannot say before November 3.
The following figures are a limited, manually reviewed set of public polling, not a comprehensive index of every poll in each race. They do not assign win probabilities, adjust undecided voters or claim that reported survey support equals the final vote. All data are frozen to a dated editorial edition.
Explore the research
Change the rules and the displayed averages recalculate from the underlying polls. Every weight is shown below.
Democratic and Republican support shown separately; undecided/other responses are never reassigned. Changing these controls does not produce a winner projection.
Poll-by-poll evidence
Loading data…
Loading survey history…
| Pollster / sponsor | Field dates / release | Sample | Published support D / R | Share of average | Source |
|---|
Sources are original releases where available. Any third-party-only listing is excluded pending review, and the reason is shown. A missing sample size is not guessed. Visit each link to verify original wording, dates and methodology.
Show your work.
We use a basic descriptive weighted mean of the two major-party percentages from published general-election questions. This is a research tool, not an election forecasting model.
- Filter. Use surveys fielded within 45 days of the dated editorial snapshot. Require field dates, candidate figures, population type and a traceable publication. One latest eligible poll from each polling organization is included. Unverified originals and questions that use incompatible ballot formats stay visible but do not contribute.
- Weight. The default weight is
0.5^(age_in_days / 14) × clamp(√(n / 1000), 0.7, 1.35) × population_factor. Likely voters use 1.00, registered voters 0.75. Unknown sample sizes use a neutral size multiplier of 1.00, explicitly marked. Pollsters are not assigned subjective grades or hidden adjustments. - Aggregate. Calculate the weighted mean of Democratic and Republican published support independently. Do not normalize to 100%, distribute undecideds, or model turnout. Require at least three distinct pollsters; the site suppresses averages below that threshold.
- Test sensitivity. The controls let readers swap to equal pollster weights or remove registered-voter polls. Shifts when the rule changes are a reminder that modeling choices matter.
- Track changes. An editor must verify newer primary releases, update the source JSON and set a new editorial date. The live website cannot automatically claim a fresher poll check. Preserve earlier editions for auditing. In ranked-choice races, average only comparable first-choice results; never mix them with simulated final-round shares.
What this does not estimate
There are no election-win probabilities, predicted winners, vote-share forecasts, turnout adjustments, house effects or mathematical confidence intervals on this site. A poll's own reported margin of error is not an uncertainty interval for a weighted average.
Known limitations
This edition includes manually checked public sources for nine states and mixes some question types and sample populations. Pollsters may have correlated errors; nonresponse and likely-voter assumptions may introduce systematic error. Comparisons must keep fieldwork, populations and ballot wording in context.
Research context: AAPOR best practices. Methodology last revised October 26, 2026. Questions and corrections: corrections@electiontracker.live.
October 26 baseline, preserved.
Download the exact October 26 research edition. The eventual final pre-election snapshot has not been published. It requires a fresh review of newly released surveys and an editorial lock before polls open on November 3. The release package includes a validator for producing that later archive.
More ElectionTracker research
All 14 race guides · Dated election changes · Poll closing timeline · Election-night results center · Official voting resources.
No live data feed is connected to this research edition. Verify newer developments directly with the linked polling organizations.