The problem I have with the polling average is that it treats all samples sizes the same, which they are not. The intial weighing of the data should be done in accordance with sample size.
I wish for this also, although it brings up a further question about weighting of Likely Voter vs. Registered Voter (not to mention Adult) polls. An even more important issue is momentum. When there is a big shift in support, averaging in the oldest polls equally just throws everything off. They should weight based on the age of each poll.
I see a lot of complaining and disbelief about polls people don't like, but in my own attempts to develop predictions in 2008 and 2010, I've learned the hard way that the polls, properly understood within their limitations, do almost always tell the truth.
We may not like them, but about the only valid basis I have for disregarding a given poll now is another, better-conducted, poll. That is, citing particular anecdotal data or lines of argument rarely gets us closer to actual voting results, because that data is already reflected in the thinking of the electorate.
And I have problems with all of the polls that don’t weight their sample size by the appearance of conservative, moderate, liberal, and cross-over in Iowa, the US, or any state.
They weight by party when doing a head to head between the major party nominees, but they don’t reveal that they weight the sample by self-identified party niche within republican-only polls.