Backtesting is the process of testing the validity of a statistical models. Backtesting is an extremely important part of risk management, and can be used to detect theoretical, system, and operational errors.

As an example, if you are measuring 1-day 95% value at risk, then over 100 days you would expect to see 5 exceedances. Over any given 100 day period you may see a few more or a few less, but if you see no exceedance or more than 10, this may indicate that your model is no longer working properly.


Back to glossary index