Hi everybody! This thread is very interesting, so I decided to do my first post. My approach to Strategies and Portfolios optimization is trading futures, but I think all these processes can be carried into the FX markets.
I spend a lot of years looking for a perfect timing combination in Walk-Fordward tests, until I realized that no WFO can beat a pure optimization. I did my tests and I reached the same conclusión that Andrés García says in this study about In-Sample and Out-Sample (sorry, it´s in spanish, but maybe Google can help you):
You can do this tests with your own strategies… I did with all mine and the result was always the same: pure backtest will give you the best parameter combination. I only do a monthly backtest to look for changes.
I use System Quality Number (SQN) as target measure to optimize my strategies (a Van K. Tharp index that fits, in one number, Sharpe Ratio and number of trades).
You can think that a pure optimization can´t give you a proper projection of results, and you are right. I do a 99% Montecarlo-Test with all the trades and take compare this drawdown with the historical DD. Imagine a have a 20.000$ historical DD, and a 25.000$ Montecarlo DD, I take this coefficient (20.000/25.000=0.8) as a reducing coefficient on performance expectations over pure optimization test. So far it works like a charm.
The next step is to balance all your strategies in order to measure each one to fit in one portfolio that reach your desiderable numbers: Under your maximum DD (remember, it´s a Montecarlo DD), profit enough and a minimum “max time to recover” (so important). I use my own tools to do this process… it´s easy but it takes a lot of work (and it´s in continuous improving).
My strategies only trade during the session (no overnight positions open), and I use 5 minutes or higher timeframes. The rules are very simple (I don´t use estándar indicators, only mine). All strategies have time filters and only trade in one direction (buy or sell, never both).
I take between 10-12 years of historical data to optimize (it depends on available).
One important thing to finish. How do you know if a strategy is no longer working as it must? You have to monitorize each strategy incurring in DD. I use Fail system Montecarlo detection tool of MSA (Market System Analyzer). It´s great to know if an estrategy is in its historical profit/loss bounds, or if it´s out of control.
There are a lot of processes hard to explain, and maybe you must do your own ones, that fit best with your trading style. Only experience can tells you the way!