Calibration trainer
Being right matters, but so does knowing how likely you are to be right. A well-calibrated person is right about 70% of the time when they say 70%. Almost nobody starts out that way, and the only way to find out is to keep score over many answers.
130 questions · keys 1 and 2 to answer, 5 to 0 for confidence · enter to lock in
10 questions with two answers each. Pick one, then say how sure you are. Being wrong costs little at 60% and a lot at 99%, so the best score comes from saying exactly as sure as you are.
- Best
- …
- Answers
- …
- Streak
- …
Your calibration
Loading your answers…
| You said | Answers | Right | Gap |
|---|---|---|---|
| 50% | 0 | n/a | |
| 60% | 0 | n/a | |
| 70% | 0 | n/a | |
| 80% | 0 | n/a | |
| 90% | 0 | n/a | |
| 99% | 0 | n/a |
How to read it
Each dot gathers every answer you gave at one confidence level. Across is what you said, up is how often you were right. If your 80% answers come out right 80% of the time, the dot sits on the dashed line. Dots below the line mean you were surer than you had reason to be; dots above it mean you knew more than you gave yourself credit for. Bigger dots hold more answers.
Most people start out overconfident on general-knowledge questions like these, and most of all on the hard ones, as Lichtenstein, Fischhoff and Phillips found when they reviewed the research in 1982. The fix is not to doubt everything. It is to notice which kinds of question fool you and say 60% or 70% on those. The Brier score rewards exactly that. It is the squared gap between what you said and what happened: a miss at 99% costs 0.98, twice the 0.49 of a miss at 70%, while being right at 99% instead of 70% saves only 0.09.