Or, if trading several instruments (a portfolio) simultaneously, drawing portfolio-wide returns from the return set (i.e. the equity curve) to produce synthetic account histories. You can then calculate the standard collection of trader's statistics on each of these histories, and build a probability density function for max drawdown or Sharpe Ratio or %Winning Months or whatever statistic happens to float your boat.
Options Futures and Other Derivatives by Hull Lots of detail of Brownian motion and how to establish a Monte Carlo simulation. If you do not want to buy it then go to Borders with a notebook and jot it down.
So let me get this straight: Letâs say you have 120 months of historical data. We want to evaluate monthly profit and max drawdown. You take random 20 months of the historical data (that is you donât generate random price path) and then you apply your strategy. You repeat this n times. Is this correct? Thanks
This approach makes a very important, and potentially expensive, assumption. Specifically, that future performance will adequately resemble the historic probability distribution of outcomes. I think that people would do well to have a large margin for error in place to ensure long-term survival. What irks me is when people use MC and then discuss strategy to the second decimal point, thinking that they can rely on that level of accuracy.
Go to www.palisade.com and get @Risk it has everything you need, unless you want Levy Flight simulations which means you will need to code them yourself. One other tip always use the latin hyper cube sampling not the monte carlo!
Yes it requires stationarity. This is one of the main differences between playing gambling games where the rules of the game tend to provide fixed statistical properties and trading in markets where the statistics change. Testing for stationarity is a very important step in evaluating trading systems.
Yes, I agree, dont buy the book. You dont want get a book like that too close to the other books in your library. Steve
Other downsides of Monte Carlo is that is assumes constant volatility and has an upward drift. We all are very aware that volatility is never constant and the markets do not always go up. Steve: Hull is good to have for show in the bookshelf lol....We used it in school and is a great read if your first language is formulas,
http://www.rotman.utoronto.ca/~hull/ His course material is on there if anyone is interested. It's not in detail. But it got me thinking a few months ago when I went over the material. ozzz