Value at Risk

Value at Risk measures the loss that occurs at a confidence level. The typical confidence levels are 90%, 95%, and 99%. At each level, a certain loss is given based on either the parametric method, the historical method, or the fat-tailed method of calculating Value at Risk.

The parametric value at risk uses normal distribution and has all the issues that come with assuming a normal distribution. It can overestimate upside return and underestimate downside return, which is pretty important if you’re trying to determine the losses you’re likely to suffer. This is one of the reasons I prefer to use historical value at risk. Historical value at risk simply uses actual real values for the 90%, 95%, and 99% returns as Value at Risk numbers. The lowest returns are selected if sorted from highs to lows. Note that there are more highly complex models for value at risk that are also more accurate. These are very simple ways of calculating average true range.