January 3rd, 2007, 2:51 am
And also be aware that in some situations a parametric delta-gamma approach is not going to work. Suppose you made a volatility trade (straddle), your VaR risk will be underestimated. My preferred approach is a multivariate Monte Carlo simulation simulating the spot underlying (capturing delta and gamma risk) and the volatility surface (delta/time vertices) in order to caputure Vega/Smile/Skew Risk. Read carefully 'Delta-Gamma four ways' because, if I´m not wrong, they mention some problematic situations using Cornish-Fisher approximation.Best regards