Strategy testing and simulation

MCJ,

Those TradingBlox graphics are great, and your system looks quite solid. The problem with that Monte-Carlo simulation is that it (normally) relies on the Geometric Brownian Motion (GBM) model to drive it's simulations. The GBM makes two important and quite false assumptions: (1) Returns are independent from one period to the next (one period's return in no way affects the next's) and (2) they are identically distributed (volatility never changes). This is otherwise called the iid (independent and identically distributed) assumptions and they are obviously false (but schools and institutes go on teaching it). I've attached two graphs showing returns on a real stock and returns on a GBM simulated stock. You can see that the GBM simulation in no way resembles the real data. It is normally distributed and the real data has fat tails. Have you investigated other models that don't make these assumptions for use in your optimizations and simulations?

my.php

my.php


Gyles,

Nice surfaces, but I'm a bigger fan of R graphics. The learning curve is worth it if you're get into quantitative finance.

http://addictedtor.free.fr/graphiques/thumbs.php


Mud
 
You must log in or register to reply here.
Back
Top