Chance of each outcome across 1,000 simulated counts.
The ABC's election coverage, side by side with the projection. YouTube runs 20–60 seconds behind TV.
No sound? Tap the video, then the speaker icon. On iPhone, use the volume buttons.
88 seats coloured by who's ahead. 45 to govern. Seat numbers above count who's ahead in each seat; the small range is the 80% band across 1,000 simulated counts.
Before 6pm this shows the poll average. Once booths report, it shows what the count implies, and how far the polls were off.
Most election-night coverage measures the swing from the last election. This page asks a different question: how wrong were the polls?
Before a single vote is counted, every seat starts where the polls put it. That's the average of the published polls (Coalition 30.6, Labor 22.3, One Nation 22.0, Greens 14.4), applied to each seat's 2022 result. Preferences follow the real flows seen at the 2026 South Australian election and the Farrer by-election.
As each booth reports, it's compared with the same booth in 2022, and that swing is compared with the swing the polls predicted. The gap is the polls' error. It updates every seat at once, including seats with no results yet. Each seat's chances come from 1,000 simulated counts.
Early numbers jump around. Small country booths report first, and the big pre-poll centres (most of the vote this year) arrive late in the evening.
An experiment by Farrelly & the Machine.