D is correct. Since historical simulation requires no parametric assumptions about a distribution of P/L, it can easily accommodate fat tails and skewness.

A is incorrect. A major potential weakness of a historical simulation approach is that its results are very dependent on the historical data set. If the period these data were taken from was unusually quiet or unusually volatile (or is otherwise not reflective of likely future events), the ES estimate could fail to accurately represent the risk we are actually facing, and a well-specified parametric estimation method could be more accurate.

B is incorrect. There is a tradeoff between these two objectives when choosing the length of the sample window for a historical simulation. A reasonably long window is needed for the sample size to be large enough to get risk estimates of acceptable precision; however, the longer the window, the more the news in current market observations is likely to be drowned out by older observations.

C is incorrect. A historical simulation approach can have difficulty handling shifts that take place during the sample period. It can be slow to reflect permanent shifts in a market's risk profile or episodic changes in risk resulting from major events.