Playing with opinion polls

Now assuming that voting intention polls actually sample properly and there is no Shy Tory Factor and Shy Labour Factor and assuming a smooth national swing (I know I should do it on a constituency basis) then averaging all of the polls for the week coming up to the election from all of the pollsters and using the BBC uniform swing predictor this gives the following results;

Cons: 34.2% CI 33.6% - 34.8%
Lab: 27.2% CI 26.5% - 27.9%
Lib-Dem: 28.6% CI 27.9% - 29.3%

This equates to the ranges of seat numbers based of the other parties getting minimum share when one party gets maximum share.

Cons: 263-285
Lab: 249-273
Lib-Dem: 85-99
Others: 28

So you cannot predict who will be the largest party on the poll data