Simulation is a great way to understand a system if you can model the stochastic data correctly. Using a distribution based off of real data could simulate in a way that you could run long run probabilities and be significantly confident in the results, as long as the underlying fundamental market conditions dont change. If they do then you have no basis for the simulation to run from that is representative of the market. I also would not recommend it for short runs, just as I wouldnt run a back test using short runs.