Skip to content

Strict and inclusive VaR exceedance

At a discrete VaR boundary, strict and inclusive loss events can have different probabilities.

On this page

Use the same threshold

In the original loss model, the lower 90% quantile is 10. Loss 50 carries 5% probability and loss 10 carries another 25%. The definition of the event determines whether that boundary mass is included.

Compare the events

Original event probabilities
EventIncluded outcomesProbability
L > 10505%
L ≥ 1010 and 5030%
L ≤ 10−10, 0 and 1095%

Pair the correct complements

The complement of L>10 is L≤10. The complement of L≥10 is L<10. Mixing those pairs produces a probability discrepancy equal to the mass at the threshold.

Do not infer ES from the event label alone

A selected exact-tail average may use part of the boundary mass. The tool reports both strict and inclusive probabilities so readers can see why an unqualified exceedance statement would be incomplete.

Further references