In the study notes it is mentioned that a good model will produce approx. the number of expected exceptions and an example of 95% VaR model is used over 250 days. The computation shows that it will produce approximately 5% * 250 days = 25 days exception. Is that a typo? Should it not be 12.5 days. A few lines below however, the correct computation of 12.5 days seems to be mentioned.
Hi @NAndr5521 Yes, sounds like a typo. We expect a good (aka, accurate) VaR model to generate (1-confidence)% exceptions under the typical (i.i.d.) assumption such that we expect a good 95.0% VaR to generate 12.5 exceptions over 250 days, or 25.0 exceptions over 500 days. This is because (if i.i.d.) it's a binomial distribution with mean of p*T. Hope that's helpful!
Similar threads
