VAR attempts to assess the odds of losing money on a portfolio of, say, shares. There are three key inputs.
First, a confidence level – nothing about the future is ever known with absolute certainty, but VAR tries to offer 95% or even 99% confidence. Next, there’s a time period – this could be a day, week or month. Finally, there’s an estimated worst-case loss.
The actual calculation is complex and can be done several different ways. But the conclusion is typically similar, for example, “I am 99% confident that if I invest $1m now I will not lose more than 7% of it by the end of the day”.
Sadly, there is always a small chance of a random ‘black swan’ event occurring (such as a stockmarket crash). If your VAR model fails to build that in, you could end up losing a lot more.