Here is a discussion of a poll aggregation methodology which worked extraordinarily well:
http://www.huffingtonpost.com/simon-jackman/modelbased-poll-averaging_b_1883525.html You can't just add up poll numbers and take the average. You have to take into account sample sizes, the date of the poll, whether it was a single day or a rolling average, and a bias or "house effect" for a given pollster. - Jed