Federal Data

1. The polling average

The federal average uses the 10 most recent national polls (currently 2 Oct 2026 to 29 Oct 2026). Each poll is weighted by the square root of its sample size (1,000 where none is published) and by how recent it is: its weight halves for every 21 days it is older than the newest poll. Each party's figure is the weighted mean of the polls that report it:

weight = √sample × 0.5^{(days older ÷ 21)}  ·  average = Σ(weight × result) ÷ Σ weight

The two-party-preferred figures (Labor v Coalition and Labor v One Nation) are averaged the same way, from the 10 most recent polls that publish each one. Turning a pollster off above removes its polls and the next most recent poll takes its place.

2. If an election were held today

• Starting point: each of the 150 divisions starts from its primary votes at the 2025 election, except where a by-election has been held since (Farrer): there it starts from the by-election result, moved by the change in polling since polling day, so a new member is judged on the vote that elected them. A major party that did not stand at the by-election keeps its 2025 share in that seat.
• Swing: the change in each party's average since 2025 is added to every seat (uniform swing). Independents and minor parties are scaled up or down with the national Others vote instead.
• One Nation: because One Nation is up sharply, its seat-level vote is set from the national figure times each seat's propensity: higher where the Greens vote was low and where One Nation already polled well. The other parties share what is left in proportion to their vote in that seat, so its gains come from Labor in Labor seats and from the Coalition in Coalition seats. This is blended in fully once One Nation is 10 points up on its last result, and was fitted on the 2022–2026 South Australian surge.
• Preferences: candidates are excluded from the bottom and their votes passed on with preference flows tuned so the national two-party counts they imply match the polled 2PPs. Where the final two are Labor and the Coalition, the result is averaged with the electoral pendulum: the seat's 2025 2PP moved by the national 2PP swing.
• Uncertainty: the whole count is repeated 1,000 times with random error: about ±1.5 points nationally for Labor and the Coalition, ±1.2 for the Greens, ±2 for One Nation, ±10% on Others, ±1.5 on each 2PP, and ±2.5 points on each candidate in each seat. The likely range is the middle 80% of runs; a seat's chance is the share of runs its leader wins.

3. The seat map

On the Interactive map tab, each division is coloured by the party that wins it in the central estimate (no random error). Its card names the projected winner as a GAIN (from whoever holds the seat now) or a RETAIN (starred, with a note, where a by-election or a change of party has moved it since 2025), then shows the estimated primary vote, the final two-candidate count and the chance of each party winning across the 1,000 runs. Coalition wins in seats held by the Nationals are shown in Nationals green.

4. How reliable is it?

Tested on the 2025 federal, 2024 Queensland and 2026 South Australian elections using the final polls, the model called about 9 seats in 10 correctly, and it projected the 4 seats One Nation won in South Australia in 2026. It cannot see local factors such as a retiring member, a strong local candidate or a campaign concentrated in a few seats, so individual seats are less certain than the overall count. Published MRP projections are listed separately under Seat projections.