FRM Part II · Market Risk · Free Lesson

Backtesting VaR and VaR Mapping

Free GARP FRM Part II lesson in Market Risk. 21 min read, ~3,100 words.

A trading desk's 99% one-day VaR breached eight times last year on 250 trading days. The model claimed 2.5 expected breaches. The supervisor wants to know whether the model is broken or you got unlucky. Kupiec gives one answer, Christoffersen gives another, and Basel's traffic light tells you what your capital surcharge will be. Three frameworks; the exam tests all three.

A VaR model that nobody checks is a number with no meaning. Regulators require backtesting because risk models can drift away from reality silently: vol regimes shift, correlations move, fat tails appear. The exception-counting framework is the simplest possible check: a 99% one-day VaR should be exceeded 1% of the time. Count the exceptions; compare to expectation.

KEY: Backtesting gives you statistical evidence about whether a VaR model deserves to be trusted with capital decisions. Without it, the VaR number is hope, not measurement.

Kupiec's proportion-of-failures (POF) test is the standard exception-count check. The null hypothesis: the breach rate equals the model's stated p (e.g., 1% for a 99% VaR).

Read the full lesson, free →
Worked examples and practice. Free with a free account, no card.

Common mistakes

Bottom line

Exam shortcut

When a question gives a breach count and asks for action, the Basel traffic light decides: ≤4 green, 5-9 yellow, 10+ red. If the question asks whether a model is rejected, run Kupiec and compare LR to 3.84 (1 df). When mapping a bond, the cash-flow approach is the most defensible answer; principal mapping is an exam trap that ignores key-rate risk.

The full lesson (about 3,100 words, 21 min read) adds 2 worked examples, all 6 common mistakes, a self-check, free in the app.

Learning objectives

Browse all free FRM Part II lessons or jump into free FRM Part II practice questions.