Skip to main content Skip to main content

Calibration

The match between expressed confidence and realized accuracy across a sequence of forecasts. A well-calibrated analyst who expresses 70% confidence is correct roughly 70% of the time across the population of 70%-confidence calls; a systematically overconfident analyst expresses 70% confidence but is correct only 50% of the time. Calibration is measured by bucketing logged decisions by stated confidence and comparing the bucket midpoint to the realized hit rate. The annual calibration curve is the most honest scorecard a practitioner can produce and is the single most actionable feedback the decision journal generates.

Lessons that use this term

Related terms

Ambiguity Aversion · Anchored Assumption · Asset Beta · Bank ROE Spread · Banker Pitch Deck · Beta

Open this term in the app → — no account needed; browse the full glossary while you research.