Best for: Analysts, investors, leaders and anyone who wants to predict better.
1 min read · 3 lessons
Summary
What the book is about
The big idea
Good forecasting is a learnable skill, not a gift.
Philip Tetlock spent years studying expert predictions and found many experts did little better than chance. Then he ran the Good Judgment Project, part of a forecasting tournament funded by a US intelligence research agency.
A small group of ordinary volunteers, which he called superforecasters, consistently beat others, including some intelligence analysts with access to classified information.
With journalist Dan Gardner, Tetlock explains what made them so good: their habits of mind, not genius.
Key takeaways
The ideas worth keeping
01
Forecasting is a skill.
Good forecasters improve through practice and feedback. Tetlock found that people who kept score and studied their misses improved year after year.
02
Break problems down.
Split big questions into smaller, answerable parts. Fermi-style estimates turn a scary unknown into a chain of smaller guesses you can reason about.
03
Start with base rates.
Begin with how often similar events happen, then adjust for specifics. This is the outside view: what usually happens to cases like this one, before you look at the details.
04
Update incrementally.
Change your mind often, but in small steps, as new information comes in. Big swings in opinion are a warning sign. Steady, small corrections track reality better.
05
Be precise.
Use specific probabilities rather than vague words like likely. Saying 70 percent instead of 'probably' lets you be checked, and being checked is how you get better.
06
Keep score and reflect.
Track forecasts, measure accuracy and learn from mistakes. Superforecasters ran honest postmortems on their hits and misses alike, asking what they got right for the wrong reasons.
3 lessons
Put it into practice
- 01
Break big questions into small ones.
Superforecasters split impossible questions into pieces they can estimate, the way physicist Enrico Fermi did.
- 02
Update often and in small steps.
The best forecasters change their minds frequently, but by small amounts as evidence arrives.
- 03
Keep score.
You cannot improve what you do not measure. Write predictions down with probabilities and check them later.