Model Risk Directory emblemModel Risk Directory
Validation Techniques

Backtesting

A statistical technique that compares a model's historical predictions against what actually occurred, used to assess predictive accuracy over time. Backtesting is a specific, quantitative form of outcomes analysis, common in credit scoring, market risk (e.g. Value-at-Risk), and capital models.

Last updated: Last reviewed by: Model Risk Directory editorial team

How the term is used in model risk management

A statistical technique that compares a model's historical predictions against what actually occurred, used to assess predictive accuracy over time. Backtesting is a specific, quantitative form of outcomes analysis, common in credit scoring, market risk (e.g. Value-at-Risk), and capital models. The exact implementation varies by institution, model type, risk rating, and governing framework. Use the linked regulatory pages and buyer guides below for scope-specific requirements.

Related glossary terms

Outcomes analysisModel validationOngoing monitoring

Guides and frameworks that use this term

GuideModel Risk ValidationModel risk validation is the independent testing of a specific model's conceptual soundness, ongoing performance, and outcomes, performed by staff or a third-party firm genuinely separate from that model's development. It is the 'effective challenge' function required under SR 26-2 and SR 11-7, distinct from broader model risk management program design.GuideTop model risk management firmsThe right model risk management firm depends on the models in scope, governing framework, independence requirements, technical methods, jurisdiction, and delivery capacity. This guide groups the directory's advisory firms alphabetically and provides a verification checklist. It does not assign rankings, scores, or endorsements without comparable evidence.FrameworkHKMA CA-G-4HKMA CA-G-4 sets validation expectations for internal-ratings-based credit-risk systems used by authorized institutions. The current 2025 version covers governance, data, model design, quantitative and qualitative validation, use tests, stress testing, overrides, validation frequency, independence, findings, and supervisory review. It is model-specific rather than an enterprise-wide MRM standard.