Uncertainty — Federal

Seat ranges

The number on the forecast page is the centre of a distribution, not a flat prediction. Across 50,000 simulations, the 80% interval contains the outcome eight times in ten — that is the range to quote. The 95% interval is wider and more cautious; it shows how far the model concedes the race could go.

Width — how wide the 80% range is — is the most honest single measure of a party’s uncertainty: a party whose range spans sixty seats is not projected, it is estimated.

Where each party lands

Finishing first and forming a government are two questions; finishing second is a third, and it decides the official opposition. Each bar splits a party’s simulations across those outcomes.

Convention: on a tie at the top, both parties count as first, and that simulation counts a second place for nobody.

Projection trajectory

Each point is a full model run published that day, not a poll. The line shows how OUR read of the race moved — under incoming polls, but also under model corrections.