Backtesting

Roles that need it

• Model Validation Analyst
• Market Risk Analyst
• Credit Risk Model Developer
• Investment Risk Analyst

Skills it belongs to

• Model validation and governance
• Hypothesis testing and significance
• Descriptive statistics and distributions

Related concepts

• Value at risk (VaR)Value at risk estimates the loss a portfolio should not exceed over a given horizon at a given confidence level, for example 99 percent over one day. Common methods are historical simulation, parametric (variance-covariance) and Monte Carlo. It says nothing about the size of losses beyond the threshold.
• Model validationModel validation is the independent review that checks whether a model is conceptually sound, built on suitable data, correctly implemented and performing as intended for its use. It is done before first use and periodically after, and its findings feed model approval and the model risk framework.
• Probability of default (PD)Probability of default is the likelihood that a borrower defaults over a given horizon, usually one year. Capital models use a long-run average (through-the-cycle) PD per rating grade, while IFRS 9 uses point-in-time 12-month and lifetime PDs that react to the economic outlook.
• Discriminatory power (Gini, AUC)Discriminatory power measures how well a rating model or scorecard ranks borrowers who later default above those who do not. Common statistics are the area under the ROC curve (AUC) and the Gini coefficient, or accuracy ratio, equal to 2 x AUC minus 1.

Sources

• Guidelines on PD estimation, LGD estimation and treatment of defaulted assets, EBA (external site) accessed 31 Oct 2026
• SR 11-7 Guidance on Model Risk Management, Federal Reserve (external site) accessed 31 Oct 2026

Updated 31 Oct 2026.