"The Signal and the Noise" by Nate Silver, ISBN 978-1594204111
Excellent book all around, highly recommended.
Nate Silver is famous for correctly predicting almost all house and senate races in the US elections. Previously, he has correctly predicted a lot of pro sports outcomes (mostly baseball). In this engaging and easy to read book he explains his work predicting weather, politics, sports, gambling and stock market.
Major themes:
- prediction is hard
- averaging multiple predictions beats any individual one
- weather forecasting became much better in the last 50 years, while financial market forecasting remains the same
Nate writes the simplest explanation of Bayesian theory I have ever read. For this alone he deserves highest praise. Later he applies it to real life situations, and while some are sad (spouse cheating), others are very interesting (poker).
If there was single bone to pick with this book, it would be his reliance on pie charts. They are aweful, since humans are terrible at comparing areas of circles and arcs. Especially bad are several pie charts side by side with different radii, showing different ratios. How one is supposed to compare sector values is besides me.
Bonus: good explanation of what happended during Garry Kasparov's vs Deep Blue chess match.
Application to software development
- Time estimates: the book essentially argues that agile model's estimation "poker", when everyone on the team estimates every task followed by a trading round, is a good strategy.
- Human and process improvement follow 80-20 rule. Any good software process is necessary, but to beat the competion you need to reach the remaining 20% by expanding more and more energy.
- Stop trying to estimate what would happen. Instead prepare for a variety of possible outcomes. Thus investment in software quality would pay off, because you make it possible to quickly pivot depending on the unpredictable feedback.