Thread by @kevinbonham

Just a note on what happened with sites doing projections on Saturday night that resulted in a number of apparent Labor wins and what seemed a near-certain majority collapsing. The main issue was overconfident assumptions about prepoll swings replicating booth swings. #NSWvotes

Not only is there a general trend of prepolls having a lower swing than booths at this election but the gap happens to have been especially large in some seats where Labor needed it not to be. Holsworthy 5.8% different, Terrigal 4.0%, Ryde 4.4%, Goulburn 3.3% Kiama 3.5%

This makes a difference because the number of prepolls is so huge. In Terrigal there are over 12000 prepolls. If those had swung as much as the booths then Labor would be almost 800 ahead instead of over 200 behind.

Something like this also happened in Vic 2018 but it just turned a massive thrashing into a very big thrashing so apart from red faces about the odd seat or two nobody really cared.

The other issue with some projections is the massive increase in postals.

What caused the prepoll difference? At this stage who knows. Easy to point to the cluster of polls at 53 and Newspoll at 54.5 and say late swing and maybe that's true but Newspoll was in the field day 2 of prepoll and out before the final day of it.

Indeed Freshwater polled only slightly earlier than Newspoll and had 53, people can say oh maybe there was a big swing that started on Mar 22 but this is placing too much reliance on polls that are not usually that spot-on.

The postcount shift back to the Coalition is also affecting the polling picture. On the weekend looked like Newspoll nailed it and the others pretty good but a bit off. Now all polls looking very good (Morgan primaries a bit off) and the averages even better.