There are different techniques companies use to calculate VAR. At CLOUDRISK we utilise both historical and Monte Carlo approaches to VAR.
For both methods, we use a full re-valuation of the trades for each scenario and do not use any approximation. Therefore even the most complex derivatives are captured accurately.
With the historical method, VAR is determined by taking historical data to assess the impact of market moves on a portfolio. It is then used to generate a distribution of returns on the portfolio. Because historical simulation uses real data, it can capture unexpected events and correlations that would not necessarily be captured by the other approaches. The historical approach is non-parametric which makes him convenient and reliable. The Monte Carlo method simulates numerous scenarios for the portfolio and determines VAR by observing the distribution of the resulting paths. We do not use any kind of parametric and approximation based methodologies as they have proven to perform poorly in the context of derivatives.