[Book Review] "The Signal and the Noise: Why So Many Predictions Fail—But Some Don’t.", by Nate Silver

by Douglas_Reay1 min read7th Oct 20128 comments

16

Book ReviewsForecasting & Prediction
Personal Blog

Here's a link to a review, by The Economist, of a book about prediction, some of the common ways in which people make mistakes and some of the methods by which they could improve:

Looking ahead : How to look ahead—and get it right

One paragraph from that review:

A guiding light for Mr Silver is Thomas Bayes, an 18th-century English churchman and pioneer of probability theory. Uncertainty and subjectivity are inevitable, says Mr Silver. People should not get hung up on this, and instead think about the future the way gamblers do: “as speckles of probability”. In one surprising chapter, poker, a game from which Mr Silver once earned a living, emerges as a powerful teacher of the virtues of humility and patience.

8 comments, sorted by Highlighting new comments since Today at 12:04 PM
New Comment

Has anyone here read the book? What did you think of it? Would you recommend it?

I've read it - it's inspired a short series I shall be posting shortly.

ETA: not a chapter-by-chapter review, though I might do that too; rather, I focus on one small aspect (the "waterline" model discussed in connection with the poker bubble) and use it to illustrate something else (tips and tricks in forecasting).

I read it - I think the way I described it to the Less Wrong DC group was, "like an introductory Less Wrong article, except a book."

To elaborate a bit:

  • Pro: The writing is clear and engaging.
  • Pro: There's a lot of good case-study material and a bit of useful theory connecting them.
  • Con: The book is lacking in technical detail regarding Bayesian thinking (and his formulation of Bayes' Theorem is kinda messy).
  • Con: Nate Silver is more prone to claim successful predictors as being Bayesian thinkers than he justifies in the text.
  • Con: There are a few glaring typos in the first printing (e.g. a table on which the "Correlated" and "Uncorrelated" column headers are switched).

It's easy to read, and it's worth reading for the case studies, but I'd probably put it at a "Borrow" rather than a "Buy" recommendation.

ETA: The review you posted is good.

[-][anonymous]9y 0

I've read and enjoyed a few chapters. One interesting theme is about why we have success predicting some things but not others. E.g., our ability to predict weather has actually improved in the past few decades, but there's been practically no progress for earthquakes.

There's relatively little about political prediction. I read his blog and enjoy that too, but I liked the broader perspective here.

The current post is a nice compendium of biases.

It's also over a year old - suggesting the site has gone dormant...

sigh

On the other hand, it's not as though Bias 101 changes all that rapidly.