Subscribe to DSC Newsletter

Nate Silver's famous run of successful predictions came to an halt

This is a classic. A guy who correctly predicted election results in all 50 states, and many other correct predictions, now fails.

Nate Silver

First, Nate is well known not because of his previous correct predictions, but because he got hired by the Times magazine where he contributed as a writer. Many other people got the election results correct for the 50 states, but only Nate is known for his performance. It's actually an easy prediction as most states have well known political climates (CA is liberal, TX republican). Only 3 or 4 states are difficult to predict, and you can get them right with a bit of chance (say 1 in 4 chances to get them all right).

Here's an analogy. Let's say that 8 million stock traders make totally random trades, with 50% chances to win or lose on each trade. About 8 traders will make correct trades 20 times in a row (and that's a challenge far more complicated than predicting state elections). Guess what? 4 of the 8 successful traders (out of 8 million) will get their trade wrong on day 21, and only 1 will still be correct on day 23. By day 24, it is guaranteed (by the law of probabilities) that even the "best" traders will now perform miserably (that is, no better than the "worst" traders from the first 20 days who got all their trades wrong), moving forward.

I'm sure Nate made enough money that he does not care, and he will sure very easily find new assignments. What happened to all Princeton math PhD's who run Wall Street to nearly into the ground in 2008? I'm sure most of them are still doing very well, though I've heard many switched from doing stock market arbitrage to advertising arbitrage (both require good data science expertize), not because they were fired for incompetence, but because the industry (Wall Street finances) shrunk significantly.

For those interested in the details, Silver's team wrote, about global warming, that "When you read that the cost of disasters is increasing, it's tempting to think that it must be because more storms are happening. They're not ... In reality, the numbers reflect more damage from catastrophes because the world is getting wealthier"

This is wrong, if you look at the following picture:

Munich Re Disaster Frequency and Losses, 1980-2010

There's no more earthquakes, but more floods and extreme temperature. But I'll go one step further: these charts do not prove global warming, they just prove climate change. And cost of disasters increased mostly because an increase in number and severity of events, and to a lesser extent, because we are building more and more in areas hit by floods and hurricanes, and overall population growth.

Other Links 

Views: 5739

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Comment by Vincent Granville on March 31, 2014 at 9:29am

I think it reflects bad hiring more than wrong causation analysis. And maybe he did it on purpose, maybe he's paid by someone (his employer) to debunk the global warming theory. Or maybe he genuinely does not believe in global warming (I don't either, I believe in climate change instead, with more extreme high/low temperatures and increased duration of these hot/cold events).

But attributing increased costs of damages (paid by insurance companies) to more expensive buildings, and discarding increased frequency and intensity of bad weather events, is definitely wrong. The costs are increasing even after adjusting for (real estate) inflation.

Videos

  • Add Videos
  • View All

Follow Us

© 2018   Data Science Central ®   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service