Home / Glossary / VaR Backtest

VaR Backtest

Validates the VaR model by counting historical breaches.

VaR backtesting compares actual losses against the VaR threshold to validate model accuracy. A 95% VaR should be breached about 5% of the time. If breaches occur significantly more often, the VaR model is underestimating risk. The Kupiec test provides a statistical measure of whether the breach rate is acceptable.

Formula
Breach Rate=Days with Loss > VaRTotal Days\text{Breach Rate} = \frac{\text{Days with Loss > VaR}}{\text{Total Days}}

Related Terms

Value at Risk (VaR)

The maximum expected loss at a given confidence level — but doesn't tell you how bad the tail is.

CVaR / Expected Shortfall

The average loss in the worst-case scenarios beyond VaR.

Monte Carlo Simulation

Generating thousands of possible future scenarios through random sampling.