Calibration measures how well your confidence levels match reality. A perfectly calibrated person who says "I'm 80% sure" on 100 questions should get roughly 80 correct.
Most people are systematically miscalibrated. Overconfidence (confidence > accuracy) is the dominant pattern — studies show that general-knowledge answers people feel near-certain about are wrong far more often than their confidence implies (Fischhoff et al., 1977).
Calibration is trainable. Research in the Good Judgment Project showed that forecasters given under an hour of probability training improved their accuracy (Brier scores) by about 6–11% compared with a control group (Mellers et al., 2014). MindFrame's calibration challenges directly target this skill through repeated practice with immediate, precise feedback on your confidence error.