Backtesting
FRM Part 1 glossary · Valuation And Risk Models
Definition
The statistical comparison of a risk model's predictions against realized outcomes—most commonly counting the days a portfolio loss exceeded its VaR (exceptions) to validate the model's calibration.
Backtesting in plain English
Backtesting is checking whether a risk model actually worked by replaying it against real history and counting how often reality broke its promises. The most common version tests Value at Risk (VaR) — a daily line the model claims your loss should rarely cross — by tallying the days losses blew past it (called exceptions). If the model says 'you'll only breach this on 1 day in 100' and you breach it 8 times in 100, the model is too optimistic and needs recalibrating. It's like grading a weather forecaster: if they say 10% chance of rain but it pours half the time, their model is off.
Learn these first
Practice Backtesting
Related FRM Part 1 terms
FRM Part 1 practice questions · FRM Part 1 flashcards · FRM Part 1 study guide