https://en.wikipedia.org/wiki/Brier_scoreFinished Superforecasting – The Art and Science of Prediction this weekend. Great book, would recommend 👌
I started listening on a motorcycle ride and had a hard time with wind noise so the first 2 chapters kind of elude me, but here's what I learned 👇
Forecasting is the art of predicting what's going to happen. Based on current events, past behavior, statistics, and the gut feeling of the forecaster.
We live in a probabilistic world (as the book manages to prove) so forecasting can never be perfect.
There are 3 types of forecasting questions:
- the completely unpredictable, follows chaotic math (shape of clouds)
- the completely predictable, if I drop this it will fall (basic physics)
- the interesting questions, very hard but measurably predictable (will NK test a nuke by Dec 3rd)
To improve forecasting you have to ask questions with definitive answers and timelines.
You also have to measure accuracy 👉 The Brier score. Only with feedback can you improve.
No fuzzy language. Use percentages.
"I am 90% likely to finish this feature by Monday, assuming X and Y don't happen. If they happen, likelihood falls to 75%"
Adjust your forecast with new information.
Much of the book talks about superforecasters and how they beat normal forecasters by an order of magnitude.
- early rough guess
- always updating
- thinking in 65% vs 67%, not 60 vs 70
- consider many many competing perspectives
- but discard irrelevant data
- avoid over and under confidence
All in all Superforecasting was a great book. Many references back to Thinking Fast and Slow and System1 vs System2 thinking, you should read that too.
Thanks for the recommend @hsribei