The run of exceptionally good subsample results for the SNP was finally broken today when Ipsos-Mori actually put the Tories ahead of the SNP. That hasn't happened previously in this campaign, but it's a somewhat artificial finding because the SNP were slightly ahead (and on a very healthy 45% of the vote) before the turnout filter was applied. As ever, individual subsamples are prone to huge error and should be treated with extreme caution, but an aggregate of several subsamples might conceivably tell you something interesting. Here is what the average of results over the last seven days shows...
SCOT GOES POP POLL OF POLLS
SNP 46.6% (+0.2)
Conservatives 30.1% (+2.1)
Labour 14.9% (+0.3)
Liberal Democrats 4.3% (-2.2)
I had a bit of a dilemma with this update, because Kantar/TNS provide partial information about their Scottish subsample, but don't reveal what the turnout-weighted figures are. I decided to take the view that you just have to go with the information that's actually available, so this update is based on subsamples from seven polls - two from YouGov, one from Ipsos-Mori, one from Kantar/TNS, one from Survation, one from Panelbase and one from ICM. The GfK subsample published today (which shows an enormous lead for the SNP) is excluded because the fieldwork began more than seven days ago.
A few people mentioned being interviewed by YouGov the other day for what seemed to be a Scotland-specific poll, so if it's intended for public consumption, we may not have to read the runes from subsamples for much longer.