Large Bank Examination Workshop February 2026

Backtests • Significant discrepancies must be investigated to see if they are the result of omitted risk factors, misspecification or other errors.

• Effective, high-quality back-tests can help control model risk.

• But maintaining a rigorous separation between back-test data and data used to inform model development is not easy.

• The problem is made more difficult by the fact that, for financial markets, we are working with a limited amount of historical data.

• We are also usually building on the reported work of others.

93

Backtesting

• Issues

 How often?  Annually

 Semi-annually  Daily?

 Time horizon  Past Year

 Past Month  Past Week  Daily  Acceptable Variances?

94

47

© Global Financial Markets Institute Inc.