Why Democrats Could Do Better in November Than Everyone ThinksThe book is an elegant, intellectually omnivorous, well-researched and disciplined assertion of the need for prediction and for prognostications to be as accurate and useful as possible. He looks at the egregious failures of predictions - not considering the possibility of an air attack on Pearl Harbor because of a belief that any Japanese attack would come via sabotage by U.S. resident Japanese; not seeing the possibility of an Al Qaeda attack using hijacked airplanes to ram buildings - failures due to blind spots in imagination. He also notes that in some fields, notably in predicting major earthquakes, forecasting has not made major advances because all scientific models have so far led nowhere.
But his main thesis, grossly oversimplified, is that when predictions tend to be wrong, it is because those making them do not distinguish between the signal and the noise - between information, data, events that are necessary building blocks toward accurate prognostication and other data, information, events and commentary that may appear relevant to some but essentially lead one astray. This is a good framework for looking at the early predictions for the 2014 election.