The pro-Remain group Best for Britain have published a new seats projection based on an MRP analysis of a large number of polling interviews - similar to the method YouGov used in 2017, and that proved to be more accurate than conventional polling results. The headline figures aren't surprising at all -
Liberal Democrats 17
Plaid Cymru 4
But what's very surprising are the individual seat projections for Scotland, some of which are much better for the SNP than expected, and some of which are much worse. For example, Labour are shown as being slightly ahead in Glasgow North-East. That doesn't necessarily mean that they're projected to win the seat, just that it's too close to call, but if that's typical of what's happening in Glasgow it would be very concerning. On anything even approaching a uniform swing, Glasgow North-East ought to be an easy SNP gain. But on the other hand, the SNP are shown as being fractionally ahead of the Tories in Ochil & South Perthshire, and only slightly behind them in the likes of Banff & Buchan and Gordon. Those are better results than you'd expect if the Scottish Tories really are back up to the high 20s - although admittedly the data is probably skewed by the long spell covered by the fieldwork. But the dates can't explain why the SNP appear to be virtually level-pegging with Jo Swinson in East Dunbartonshire - that's the most striking finding of the lot.
I'm not sure whether this should cause us to reevaluate our assumptions completely, or whether it's just dud data or dud methodology. I'll be interested to see what YouGov's own analysis shows.
* * *
Click here for a handy list of SNP election crowdfunders.