Methodology

How it works. Each district starts with its 2024 presidential margin on the lines used in 2026. The generic ballot then moves every district by the same amount. Because Republicans won the 2024 national vote by 1.5 points, a D+6 ballot is a 7.5-point shift toward Democrats. Incumbents also carry a personal cushion into the district where they are running. Open seats have no cushion.

Generic ballot opening average. When the page loads, the opening ballot is pulled dynamically from the Silver Bulletin's public sheet of 2026 generic-ballot polls. The wheel uses our weighted average of the polls, with each one weighted by the sheet's influence column, rather than Silver Bulletin's smoothed trend line. If the sheet cannot be reached, the page uses the latest average baked into the build.

The cushion, over three cycles. The cushion measures how much better or worse a member ran than their party's presidential candidate. It uses results from 2024, 2022 and 2020. The newest result gets the most weight: 2024 counts fully, 2022 counts two-thirds and 2020 counts one-third. The model leaves out races with no useful two-party comparison, races the member did not enter, and races where the member beat an incumbent. It then adjusts the weights to use the results that remain. Of the 360 members running, 167 have three usable cycles, 92 have two, 71 have one and 30 have none.

Putting each cycle on the same scale. A House result reflects both the candidate and the national mood that year. Without an adjustment, the 2022 environment would make Republicans look stronger and Democrats weaker for reasons unrelated to the candidates. The model corrects for that gap: 4.9 points in 2022, 1.7 points in 2020 and no change in 2024. Each race card shows the adjusted results used for that member.

Limiting unusual results. A weak opponent can produce a landslide that says little about the winner's personal strength. The model therefore limits any one cycle to 10 points in either direction before applying the weights. This affects 45 cycles across 34 members. For example, Marcy Kaptur's 2022 result was boosted by her opponent's collapsed campaign. The limit lowers her cushion from 12.2 to 8.5 points, close to her other results. Race cards show when a result was limited.

Candidate quality. Use the slider on a race card to add an adjustment from D+10 to R+10 for that race only. This lets you test the effect of a strong or weak candidate. The adjustment is added to any cushion and penalty already in the model. It lasts while the browser tab is open, including through a reload. Clear all removes every adjustment. This is a tool for testing an assumption, not a reason to force a preferred result.

New lines reduce the cushion. Ten states redrew their maps for 2026, affecting 181 districts. Members keep less of their personal advantage when they inherit new voters. Evidence from the last redraw shows that the more a district changed, the less of its earlier result carried over. The model therefore keeps three-quarters of the cushion in every redrawn state. This is a modest estimate: it may still give too much credit to members whose districts changed the most.

First re-election. Forty-five members won open seats in 2024 and are running again. Their 2024 results did not include an incumbency advantage, so the model halves their measured result and adds 1.4 points. Past elections support both adjustments: about half of a first result carries into the next race, while a new incumbent gains about 1.4 points. For example, Jeff Hurd ran 4.8 points behind Trump in 2024 and now receives a −1.0-point cushion.

Incumbency control. Full, Half and Off change every cushion together. Use them to see how much the result depends on incumbency. A cushion can help or hurt: some members run behind their party's presidential result.

Battleground damper. Close races often respond less to a national swing because they attract more money, attention and campaigning. At 50 percent, the damper gives the ten seats closest to control of the House half of the national swing. The next ten get half of that discount. The other 415 seats absorb the difference, keeping the national total unchanged. At 100 percent, every seat gets the same swing. The bands use margins from before the damper is applied, so changing the setting does not redefine the battleground.

Random variation. A model cannot capture every event in every race. With random variation on, each seat gets one random adjustment of up to five points. Most adjustments are small; the resulting standard deviation is about 2.2 points. This represents uncertainty such as a new candidate, a retirement or an unusually effective campaign.

The page draws these adjustments once and keeps them while the tab is open. Turning random variation off shows the model alone. Turning it back on restores the same draw; reload the page for a new one. The seat total will often move by one or two because one party may hold more very close seats than the other.

A seat marked with an ∗ changes hands only because of its random adjustment. The mark appears on the wheel, cards and table. These seats should sit close to the dividing line. For a fair comparison, the page recalculates the battleground bands both with and without random variation.

Nine seats are already settled. California's top-two primary produced two candidates from the same party in nine districts. CA-04, CA-07, CA-11, CA-12, CA-14, CA-29, CA-34 and CA-37 will elect a Democrat. CA-40 will elect a Republican. These seats still count toward 435, but the swing and battleground damper cannot move them.

Crosshatching marks a pickup. On the 254 districts with unchanged lines, a hatched block means the seat changes party. A new district in a redrawn state has no previous winner, so an open-seat win is not a pickup. Defeating a sitting member still counts as one. In total, 402 of 435 seats can count as pickups.

States with new lines. Alabama, California, Florida, Louisiana, Missouri, North Carolina, Ohio, Tennessee, Texas and Utah use new maps, covering 181 districts. Virginia and South Carolina use their 2024 lines. Missouri could return to its 2022 map after a state Supreme Court appeal.

Baseline. Gains and losses are measured against a full House of 220 Republicans and 215 Democrats. Vacant seats are assigned to the party that last held them, and independent Kevin Kiley is counted with the Republican caucus. With no change in the national vote, the model gives Democrats 205 seats. The ten-seat drop from their baseline comes from the new maps.

Sources and limits. Presidential margins come from The Downballot, Dave's Redistricting App and the Texas Legislative Council. Cushion calculations use official 2020, 2022 and 2024 results from the Clerk of the House, plus presidential results by district from The Downballot and Daily Kos Elections. Incumbent and nominee information comes from Wikipedia's district tables and Ballotpedia as of August 2026. Results from safe seats may not carry well into newly competitive races. This model is a way to measure possible outcomes, not a forecast.