Value at Risk can also be computed parametrically using a method known as variance/co-variance VaR. This method allows you to simulate a range of possibilities based on historical return distribution properties rather than actual return values. It's actually more accurate even though it assumes a normal distribution especially when not much data is available