Definition
Plain language
A standard way banks estimate the worst loss they're likely to face over a set period, outside of rare extreme cases.
As stated in the literature
VaR — a risk measure giving a loss threshold that won't be exceeded with a specified probability over a horizon; a canonical upper-tail-sensitive forecasting target where distributional miscalibration matters and inverse-scaling failures bite.
Also called: VaR
Why it matters: It sets the loss thresholds firms plan and hold capital against, so a model that misjudges the worst-case tail can leave them dangerously underprepared.
For example, a bank might estimate that on 99 out of 100 days its portfolio won't lose more than a million dollars.