FiveThirtyEight blogger writes about predictions
"The Signal and the Noise: Why So Many Predictions Fail — but Some Don't" (The Penguin Press), by Nate Silver
Tuesday, October 2, 2012
Before his number-crunching polls and politics blog, FiveThirtyEight, became a must-read for political junkies, Nate Silver made thousands playing online poker and created a system for forecasting baseball player performance that was eventually bought by Baseball Prospectus.
Silver knows a bit about statistical probability.
Still, what are the odds that a stat-head, even an articulate one like Silver, could write a book on the craft of forecasting that could engage the average reader? Silver delivers an improbably breezy read on what is essentially a primer on making predictions by interlacing theory and numerical nitty-gritty with a series of stories about people who have to know numbers in "The Signal and the Noise." Readers meet a professional sports bettor, meteorologists, economists and poker champs. These are people whose livelihood depends on figuring out what score, what storm, what jobs number or what card is likely to come up next. The trick is to recognize the meaningful data — the signal — amid all the noise.
The irony is that we are awash in more data than ever — some 2.5 quintillion bytes are generated every day — but it does little good if we don't know how to use it. Silver comes down hard on economists who failed to recognize the signs of the massive financial meltdown of several years ago. They failed to understand how risky mortgage-backed securities were and how the housing crisis could trigger wider misery. The signals were there, but the models were wrong. People were overconfident. As Silver writes: "The gap between what we know and what we think we know may be widening."
Silver has even less patience for TV pundits who seem to make predictions based on entertainment value rather than careful analysis. His assessment of 733 predictions made on the McLaughlin Group found an equal chance of them being true or false. Not that it hurts the pundits' reputations. He notes that commentator Dick Morris is still on cable TV despite some very wrong predictions, like his assertion last year that Donald Trump would run for president and had a "damn good" chance of winning it.
There are some heroes in Silver's book, like Thomas Bayes, the 18th-century English mathematician who developed theorem to come up with the probability of an event.
He also praises the meteorologists who work at the National Weather Service. When they say there's a 20 percent chance of rain, it rains about 20 percent of the time. They compare favorably with some TV meteorologists who will overstate the chance of rain on the theory that it's better if viewers are treated to an unexpectedly sunny day as opposed to a rained-out picnic. In one egregious case cited by Silver, when a local weatherman in Kansas City would say there was a 100 percent chance of rain, it only rained about two-thirds of the time.
As one TV forecaster explained: Presentation takes precedence over accuracy.
More like this story
Use the comment form below to begin a discussion about this content.
Please review our Policies and Procedures before registering or commenting