D 10 5 2 Even 2 5 10 R Independent ahead Flips party
← Swipe the chart to see every seat
Swipe the chart to see every race →
Full shape: range of past misses*Dark core: middle 50%wider = more commonPolling average (center of the shape)Where this scenario landsEven: blue above = Dem/Ind ahead,red below = Republican ahead
Not up in 2026
No polling error applied.
Select a year or scenario to apply historical error, or drag a state to set its own.
Assumed bias: polling average vs. result
← toward Dem toward Rep →

Historical Polling Misses

How far each state's final polling average landed from the actual result, Senate and presidential races 2010-2024. Red means the Republican did better than the polls; blue means the Democrat (or a Democratic-aligned independent) did.

Polling average = all polls in the final 21 days. Source: Silver Bulletin raw polls, the continuation of 538's archive.

Election by election

Average miss across every state with a polled race, by how far and toward which party.

State by state

Races
Elections

The bar and colored tick show each state's average miss across the selected races; dots are the individual races ( Senate, presidential). Sorted from where polls most overstated Democrats (top) to where they most overstated Republicans. Hover a dot for details.

Every state and election at once. Each cell is the miss in points, colored by who beat the polls; hover a cell for the polling average and result. Same order as the state averages.

State Polling Averages

Silver Bulletin's polling average for every 2026 Senate race, closest first, with each poll behind it.

State

Closest races first; click a state or its bar to see its polls. The shaded bar and tick show the average, running from Even toward the leader. Dots are individual polls: solid = carries weight in Silver Bulletin's average today; faint = an older poll that no longer counts. Scale capped at 20 points. Silver Bulletin's average weights more reliable polls more heavily, adjusts for house effects and likely- vs. registered-voter samples, and updates older polls for later national trends, so it can sit away from the raw dots. Partisan-sponsored = polls in today's average conducted for a party, candidate or allied group.

All Latest Polls

Every general-election Senate poll in Silver Bulletin's averages, newest first, with pollster grades and house effects.

State
Pollster

Polls largely collected by The New York Times and averaged by Silver Bulletin. ✓ = carries weight in Silver Bulletin's current average for that race. Sponsor notes mark polls conducted for a party, candidate or allied group. Grade and house effect: Silver Bulletin's January 2026 pollster ratings (filled badge = A range); house effect is how a firm's polls have run against other polls of the same races, so R+3.6 means 3.6 points more Republican. Hover a result for candidate names.

What this is

A scenario explorer, not a forecast. It asks one question: if 2026's polls miss the way polls have missed before, who controls the Senate?

All polling averages, poll weights and pollster ratings are Silver Bulletin's. This site builds no averages or forecasts of its own: it takes Silver Bulletin's 2026 averages as published and applies past polling errors to them.

Where the numbers come from
  • 2026 polling averages and polls: Silver Bulletin, which updates them daily. This site pulls them on refresh; the time of the last pull is shown in the sidebar. Silver Bulletin draws most of its 2026 election polls from The New York Times's poll database, so the individual polls are largely collected by The Times; the averaging, weighting and adjustments are Silver Bulletin's.
  • Silver Bulletin's averages weight more reliable polls more heavily and adjust for house effects, likely- vs. registered-voter samples and later national trends. Polls conducted for a party, candidate or allied group are labeled as partisan-sponsored.
  • Pollster grades and house effects: Silver Bulletin's January 2026 pollster ratings.
  • Races without a Silver Bulletin average are treated as safe holds for the incumbent party:
How historical polling errors are measured
  • Source: Silver Bulletin's raw-polls archive (the continuation of FiveThirtyEight's), which holds every poll of each Senate and presidential general election taken in the final 21 days of the race, with the certified result. This site uses 2010–2024.
  • The miss: for each state, year and race, a simple, unweighted average of those polls' margins, minus the certified margin. Positive means the polls overstated the Democrat (or leading non-Republican); negative means they overstated the Republican.
  • What's included: a special election in the same state and year is averaged with the regular race. Runoffs, and races without exactly one Republican among the top two candidates, are left out.
  • In a scenario: a state's assumed miss is the average of its misses across the selected years and race types, each race counting once (so in a presidential year, a state with a Senate race too contributes two). A state with no polled race in those years takes the average of all selected races nationally.
  • Overall: The Historical Polling Misses tab breaks this down by year and state.
  • This calculation is this site's own, not Silver Bulletin's: a plain average of the final polls, not a weighted polling average like the ones used for 2026.
How a scenario is built
  1. Start from each race's 2026 polling average.
  2. Subtract the assumed miss. Polls as published assumes none; a preset, a single year or any mix of years uses each state's average miss in those years and race types (Senate, presidential or both).
  3. Apply the extra uniform shift, which moves every race the same number of points toward one party.
  4. The leader in the projected margin wins the seat. Margin = the leading non-Republican (a Democrat, or an independent shown as Ind+) minus the Republican, in points.

Control: 34 Democratic-caucus seats (including independents King and Sanders) and 31 Republican seats are not up. Republicans control at 50 seats, since the Vice President breaks ties; Democrats need 51 of their own. Independents running in 2026 are counted separately. If neither party reaches its mark, control is shown as split, unless the sidebar switch assumes independents caucus with Democrats.

Reading the views
  • Horseshoe: all 100 seats, with races ordered from most Democratic to most Republican. The line marks 50 seats; the shaded race is the 51st seat counted from the Democratic end, which decides control unless independents win seats. A ring marks a race that flips party.
  • Year strip: the seat count if 2026's polls miss exactly as each single year's did, shaded by who would control (a 50–50 Senate is Republican, through the Vice President's tiebreak). Click a year to apply it.
  • Map and assumed bias: the map colors each state by projected margin; the panel beside it shows how far the scenario moves each race from its polls. Hover a state in either to see the past races behind its number.
Limitations
  • Raw misses, adjusted averages. Past misses are measured on simple averages of raw polls, while Silver Bulletin's 2026 averages are adjusted. Some of a past miss may already be corrected for in today's averages, so the scenarios may overstate how far the polls move.
  • Few races per state. Each state has only a handful of past races, so its average miss partly reflects who was running: Kansas in 2014 (Roberts vs. Orman), or Maine's Susan Collins, who is on the ballot again. That may be a lasting pattern or a one-off; the hover lists let you judge.
  • The past need not repeat. Misses have run in both directions and varied from year to year. No scenario here is more likely than another.
  • Assumptions at the edges. Unpolled races are assumed safe holds, and how independents would caucus is an assumption either way.
Sources