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.