While the bias in the article is pretty clearly defending Mr. Silvers and his obvious choice, Obama, I'd say that trying to read his actual data is a bit confusing. It's not very clear where the sources of his data are from, whether they are from likely voters reporting who they might vote for, or the actual number of registered voters of a specific party, or past party votes, cold calling people from specific areas, or whatever. The fact is that each method of polling tends to accentuate specific outcomes. If you cold call people from the inner city, you'll definitely get a more Democratic outcome, if you do so in the south/rural areas you'll see a Republican outcome, etc. If you watch MSNBC or CNN then you'll see more reports containing likely voters rather than statistical votes or other such types like FOX news might use. This accentuates a bias and the media uses it to "prove" that one party is ahead and therefore it's pretty easy to see that each side attempts to win viewers/readers and also sway the vote to their favor. Honestly, with the media being in bed with certain political groups, I can't see how anyone could ever get a good quality poll. It's a crapshoot at best and the real losers are the voters who turn to biased news outlets for information.

Also, most people won't notice the details of such an endeavor, but likening the act of analyzing running statistics to trying to analyze a one-time act with possible inputs is nigh impossible. The first is a proven way to find trends in data, the latter is analogous to gambling.

Sam Wang is the most accurate, I believe, and he puts it above 90 percent. Strange that the data sets have such different interpretations, but I do believe the opinion of the statisticians that the media lie about how close it is to boost ratings. http://election.princeton.edu/