An anecdotal account without much technical materials, nonetheless an enjoyable read. Through the case studies (weather forecast, earthquake prediction, economic forecast, sports betting, poker, etc.) the readers are gradually introduced to what quantitative prediction is/should be.
If I am to distill three most important points made in the book, they would be
- Think probability; or even better, think CONDITIONAL probability, aka Bayesian.
- Giving your prediction a context, or "having a story," would result in forecasts that are much better than blindly dive into datamining.
- The evaluation of the performance of predictions should not be based (solely) on whether they got it right. Maintaining best practices is more important.
There may not be much originality in the materials, but the writing is decent and there are valuable lessons to take home. Recommended especially for people who work with quantitative data manipulation.
