The site aggregates polls and weights their results through statistical analysis of the pollsters past record and their collection methods. Fivethirtyeight's guru is Nate Silver who previously made his money and fame by analyzing baseball statistics and predicting winning teams.
We've watched the projections stay mostly even over the summer, drift blue during the Democratic Convention, and veer red after the Palin selection.
The last couple weeks have been more comfortable to watch.
Nate talks about the Bradley Effect and how he's accounted for other common poll-fooling phenomena. He builds in trends and historical shift into his projection model.
He runs 10,000 simulations based on the polling data. Deep in his blog he talks about his methodology, so you can buy into his conclusions with more than just wishful thinking.
The country will be so much better off, if Nate's analysis is correct.
Even if he isn't, at least my blood pressure will be lower for a while.