Lord Ashcroft, the arch ringmaster of the election campaign, has done it again. Out of his hat he has pulled a clutch of constituency polls that have created yet another mini news-cycle and nudged the narrative of the election in a further unexpected direction. Take your pick from his bouquet of snapshots: Charles Kennedy losing his seat of Ross, Skye & Lochaber to the SNP; Gordon Brown’s former seat of Kirkcaldy & Cowdenbeath also falling to the SNP; Jim Murphy clinging to a lead of just one point in East Renfrewshire; or Labour ahead of the Tories in High Peak, the party’s 68th target seat.

It’d be foolish to dismiss all of these polls out of hand, especially as the performance of the SNP is consistent with that detected in some of Ashcroft’s other surveys. But it’d be equally foolish to extrapolate the entire election result from tiny samples of opinion in a small number of constituencies. My issue with the size of the SNP’s advance in Scotland remains the same as ever: whether we should believe that consistency in the polls can ever translate into consistency in seats across an entire nation, based on the rather tenuous assumption that people in every seat will behave exactly the same way. We know that’s not going to be the case in England and Wales; why assume the opposite for Scotland?

Lord Ashcroft doesn’t always get things right. He got the Heywood & Middleton by-election very wrong, for example. I’m not denying that his work is extremely useful and fascinating, and is helping give us a picture of how the general election might unfold in a depth we have never enjoyed before. But I’m nervous of placing so much importance on his polls. Then again, I’m nervous of placing too much importance on any polls.