Tests · free, no sign-up
Calibration test
Ten computed quantities and a 90% range for each, then how many ranges held the truth and a calibration curve of your confidence. New questions every time.
In short
The calibration test asks for ten 90% ranges on quantities computed when you start, then shows how many of the ranges held the true value and draws your calibration curve. If you are well calibrated, about nine of the ten hold. Most first attempts hold far fewer, because the ranges are drawn too narrow for what the person actually knows.
How the test works
Ten questions, each of a different kind: a count of drawn dots, a shaded share of a grid, a product, a power, a square root, a compound sum, a count of primes, a probability. For each you type a low and a high number. The truths appear together at the end, with your range beside each, so nothing in the first nine answers can teach you how to set the tenth.
Reading the result
The headline is your hit rate: how many of ten ranges held the truth. Below seven usually means overconfidence. Ten out of ten, repeated, means ranges wider than you need. The curve reads each range as the middle 90% of a bell curve on a log scale and asks how many truths would have fallen inside at levels from 50% to 99%; a calibrated person's curve follows the diagonal. Moore and Healy distinguish this kind of overprecision from overestimating your own performance, which is a different error.
What it cannot tell you
Ten questions is a small sample. The same person can hold seven one day and nine the next, and the calibration trainer on this site exists to pile up more. Neither result says anything about knowledge or intelligence: a calibrated person can know very little, as long as their ranges say so.
Sources
- Soll and Klayman (2004), Overconfidence in interval estimates, Journal of Experimental Psychology: Learning, Memory, and Cognition
- Moore and Healy (2008), The trouble with overconfidence, Psychological Review
Text, questions and code on this page are original to MyTestAtlas. The cited studies describe the classic procedure; none of their items is reproduced.