I’ve made my prediction for the 628 seats in England, Scotland and Wales over at Anthony Wells’s site.
My prediction is:
Lab 338 inc Speaker (-65)
Con 215 (+50)
LD 65 (+14)
SNP 5 (+1)
My methodology has nothing to do with projected vote shares on Martin Baxter’s model (which everyone should have a go at, BTW). It’s based entirely on taking the current entries of Anthony’s site, ranking them in order, and taking the ones that caught my eye as being about two thirds to three quarters of the way along the range starting from the end nearer the status quo. (ie for a party with 100 seats, where punters have guessed randomly that they will win between 100 and 200 seats, my “method”, such as it is, predicts they will get 165-175.) Basically my gut feeling is that two-thirds to three quarters of people underestimate the electoral shift, and the rest overestimate it.
I note with discomfort that according to Martin’s algorithm, based on a uniform swing, my prediction requires the Tories to get at least as many votes as Labour and the Lib Dems to get over 27%. (I can’t quite get the exact figures from his site because my drop in the Labour share gives the Nationalists a couple more seats than I predict.) Obviously I think this vote share is unlikely. But tactical voting may achieve many strange things…