Regarding the Huffington Post chart above, there’s THIS:
Given the torrent of incoming data, the model now reports nearly complete certainty about Obama’s narrow leads in the most crucial tipping-point states of Iowa and Ohio, but that statistical confidence assumes that the final polls are collectively accurate and unbiased. When we factor in the historical potential for polling error, the probability of an Obama win falls to roughly 90 percent in Ohio and Iowa. An Obama win in those states is thus still very likely, but a 1-in-10 possibility still exists, given the typical historical pattern, that the polls could be wrong enough for Romney to win.
The polling snapshot in Florida, a state that Romney must carry to win, is far less clear. A new Miami Herald/El Nuevo Herald poll conducted by Mason-Dixon Polling and Research shows Romney leading by 6 percentage points (51 to 45 percent), while a new NBC/Wall Street Journal/Marist College poll gives Obama a nominal 2-point edge (49 to 47 percent). Altogether, five surveys released in Florida in the last week have given a slight edge to Obama, including two conducted by Democratic pollsters sponsored by groups supporting Obama, while three have favored Romney and one showed an exact tie.