Nate Silver on the Science of Prediction

Political junkies, economists, baseball scouts, meteorologists, and basically everyone else in the world is constantly trying to predict the future. And yet with the overwhelming amount of data that came with the information age, forecasters are often wrong – if not completely shocked – by the results.
Nate Silver  has been the outlier among the predictors. First it was his invention of a system that predicted the performance of baseball players. He followed his success in baseball by delving into politics, accurately predicting the outcome of the  2008 presidential election  in 49 of 50 states, as well as calling all 35 Senate races. His blog,  FiveThirtyEight, is now a part of The New York Times.
His new book is “The Signal and the Noise: Why So Many Predictions Fail – But Some Don’t.”  
“I think people understand that maybe the world isn’t easy to predict,” Silver says. “But some things that we think are relatively reliable are not – and vice versa.” Silver uses the example of the predictions about how many  jobs  will be created in the United States in a given week. Though these numbers are given to us as though they are precise and without a margin of error, on average they’re off by about 70,000 jobs, and sometimes as many as 200,000.  
Poll numbers, too, are often cherry-picked, depending on what narrative a certain news source is interested in telling. “If you want to make actual predictions,” Silver says, “then you have to be more careful about how you use the information presented.”
Though Nate Silver knows the danger of predictions, he does predict that President Obama will likely win the election. “If you go back through history and look at how many candidates who were ahead in the polls at this point in time, October 1, lost, it’s actually not very many at all,” Silver says. “So Romney’s odds are non-zero but also not all that great.”