Nov

27

 Nate Silver is the guy at the NYT who called the election using statistical analysis. His book The Signal and the Noise is NOT at all like I expected. It's about baseball, weather, earthquakes, poker, economics, stock markets, politics and elections.

Baseball uses statistical quantitative data (stats) and qualitative subjective data (scouts). Silver says you can quantify the subjective data by assigning it a score, or even a rank, which can be statistically analyzed. This is similar to the idea in the Predictioneer's Game, by de Mesquita, in which he takes subjective values, assigns a value and ranks them, and using the resulting spreadsheet can make statistical conclusions. Its the plus/minus pro/con column idea of decision making.

When looking at a player's potential they ask, among other things, "does he want to win enough to overcome the fear of losing." It's a good question to ask a trader as well. There are some great quotes in it. For example, "major league memory" i.e short term, forgetting of slumps. And from Bill James, "Our weaknesses and our strengths are always very intimately connected." He says in the most competitive industries, like in sports, the best forecasters must constantly innovate.

Silver discusses the advances in weather forecasting with huge computers taking up rooms. The human addition of judgment can improve forecasts by 15%, for example by incorporating the scouts impressions of players with pure statistics. Weather forecasters improve on the computer models with human judgment. They have achieved 300% increases in accuracy and can give you up to 36 hours advance notice of a storm. But after that forecasts become worthless. The time or place of Earthquakes can't be predicted, only their frequency and severity over a long periods of time.

He discusses the problems with economic forecasts with 4 million metrics retroactively adjusted. Statistics show that economic forecasts are worthless and can't tell if we are in a recession or not. He discusses poker and the use of probabilistic thinking. He espouses the rational market view which is not going to be well received in this forum.

One interesting point he makes is that aggregates of forecasters are more accurate than individual ones due to the wisdom of the masses effect. He uses this to good effect in political forecasts and polls.

All in all, I don't recommend the book. It's a aimed at a popular audience at a low conclusory level with little meat for stat oriented dailyspec members. I barely made it through it, due to lack of interest, (not inability to understand).


Comments

Name

Email

Website

Speak your mind

1 Comment so far

  1. Ed on November 30, 2012 10:28 pm

    Well, I just got my copy in the mail. I will bring it on a plane trip this weekend, which I think is the perfect time and place to be reading a non-fiction bestseller. Perhaps the fellow sitting next to me will pull out his copy too.

Archives

Resources & Links

Search