I remember being an avid read of Jay Cost during the last presidential election. Cost had his own blog and presented analyses of polls from a statistical stand point. He's now writing for Real Clear Politics and has had two very good articles about the upcoming midterm election.
Both deal with predictions about the House, the second with more statistics than the first.
It's hard to summarize a Cost article as he uses a lot of (simple) statistics. The short summary is that he feels that people are overstating the degree to which seats may change parties. If I haven't scared you off with the statistics comments, give his articles a read (especially the second one). They are well worth the read.