Survation, who used to regularly conduct Scottish polls but haven't done so for quite a while, have released their final GB-wide poll of the campaign. They announced that the Scottish sample was going to be a full 1000-strong and would be conducted as a full-scale, properly-weighted Scottish poll before being integrated into the GB poll as a subsample. We were supposed to see the full-scale Scottish numbers at some point tonight - so far that hasn't happened, but the subsample percentages have appeared in the datasets of the GB-wide poll. Unless I'm missing something, these figures will be identical to the Scottish poll, because I can't see why there would be any point in reweighting them. If I'm wrong I apologise in advance for getting people's hopes up, because the numbers are nothing short of incredible for an eve-of-election poll.
Scottish voting intentions for the general election (Survation, 10th-11th December):
SNP 46% (+5)
Conservatives 27% (+3)
Labour 15% (-7)
Liberal Democrats 10% (+2)
Seats projection (Electoral Calculus model): SNP 47 (+12), Conservatives 6 (-7), Liberal Democrats 4 (n/c), Labour 2 (-5)
The percentage changes are measured from the last full-scale Survation poll in April, so are only of limited use in judging the effect of the campaign. For example, April was before the Lib Dem surge really got underway. But Labour don't have any obvious alibi for these figures - Survation have been a particularly favourable pollster for them in the recent past. It's really surprising to see them performing 5% worse than in the YouGov MRP (particularly as the MRP suggested they had been gaining ground over the campaign).
I'm far, far more sceptical about what this poll is suggesting in respect of the SNP-Tory contest, because there's such a long history of polls overestimating the SNP's position in relation to the Tories. And by long history I mean even back into my childhood - in the 1992 election, every poll suggested that the SNP would finish second and the Tories third, but every poll was wrong. In 2017, Survation did a better job of picking up the late anti-SNP swing than any other pollster, but their final poll still overestimated the SNP's lead over the Tories by a couple of points.
The main thing that went wrong in 2017 is that people who would have voted SNP simply didn't make it to the polling stations, while Tory voters did. We've got to move heaven and earth tomorrow to make sure that doesn't happen again, no matter how cold or wet the weather is. No stone left unturned. Come on, let's do this. In 19 hours from now it'll be too late.
* * *
UPDATE: In my excitement, I missed a very obvious point here - this poll was conducted by telephone rather than the more common online method, so that might explain why the figures seem so surprising. Of course in the old days we would have automatically assumed that telephone polls are more accurate, but that question has been less clear-cut since 2016, because online polls actually performed better in the EU referendum.
* * *
UPDATE II: Unfortunately the Survation numbers reported above did prove to be inaccurate, although it's still a very good poll for the SNP. You can find the correct numbers in a fresh blogpost HERE.