judgement

Calibration - grade your maybes

The rule

Keep score of your guesses. Do the things you call "70% likely" really happen about 70 times in 100?

Where it flips

A few calls tell you very little. Luck can decide a small number. The fix is to check your hit-rate across many guesses over time, not a handful.

Keeping score is more than just right or wrong. A good guesser checks something called calibration. You gather all the calls you made at "70%" and see if about 70 in 100 came true. If your "90%" calls happen only half the time, you are too confident. Now you know, and you can fix it. This honest mirror turns guessing from mere opinion into a real skill you can improve.

A worked example

Over a year Aayra makes fifty "80% likely" calls. Only 55% happened. That is a clear sign she is too confident, so she widens her ranges. Without checking her hit-rate she would never have known. [illustrative]

How to spot it

  • ·calls grouped by stated confidence
  • ·stated odds compared to real hit-rates
  • ·overconfidence measured and corrected

Philip Tetlock & Dan Gardner · Superforecasting

Our plain-English take on Philip Tetlock & Dan Gardner’s idea, in our own words - not the book. The author is not SEBI-registered; nothing here is investment advice.