Whenever pollsters or prognosticators tout their record, they almost always point to their final predictions, and most people evaluate them on that. But most of the time they've made predictions well in advance of the elections. I realise there's more uncertainty a long way in advance from an election. But that shouldn't be an excuse to put no weighting on it whatsoever. Otherwise, the predictors could just make stuff up with impunity.
I propose that all forecasters tell us their accuracy rates at various arbitrary points before election day. They won't be all that accurate, but if they're as good as they say they should nonetheless be more accurate than other forecasters.
(I make
similar comments here).