Can a Computer Design a Trading System ? Part Six : A Twenty Instrument Portfolio

Through the past five articles of the “Can a Computer Design a Trading System?” series we have talked about many of the important aspects of genetic programming, showing several examples of what can be achieved with it and the absolute care that must be taken into account when building systems with this very powerful programming technique. I have explained to you what genetic programming is, the many problems that can arise from its inappropriate use and how it can be tremendously powerful at helping us uncover inefficiencies which are often not obvious due to our particular experience with different currency pairs. On today’s post I will share with you a very good example of the power of genetic programming — a portfolio built on 20 different currency pairs using 20 different systems.

Besides my posts on genetic programming,  during the past several months I have written a few posts about reliable control point simulations and the amazing power and flexibility they bring into the simulations of trading strategies.  Coupling the concept of reliable simulations using only EOD (end of day) data and genetic programming I decided to generate trading systems for 20 different currency pairs amongst which you could not only find all forex majors but also a large array of minors for which we have never developed any trading systems (due to the lack of long term lower time frame data).

My idea was therefore to put my genetic framework to the test by building systems that gave reliable control point simulations on the daily time frame. I chose this higher time frame because all symbols have long periods of time of EOD data and the influence of the spread and execution (which is historically extremely variable  on minors) is the least when you go to the daily charts. My first job was then to generate trading strategies for all the different currency pairs and to then use this information to generate a portfolio of trading systems.

Certainly the reliability of the systems and maintaining a strong degree of coarseness was paramount here since it would be useless to come up with a tremendously successful simulation which was only achieved through the merits of curve fitting. In order to do this I restricted all optimizations to only 2003-2009 (sadly most minors only have EOD data back to 2003) and I then performed the final backtests from 2003 to 2010 to include one year of out-of-sample testing. The optimizations where also carried out in the most coarse possible way (10% variations in the parameters as a minimum) giving a much higher confidence on the final results of the tests. The final portfolio results are very good with an average compounded yearly profit to maximum draw down ratio slightly higher than 2 and no losing years for the past 10 years (although 2008 was barely profitable).

Finally I was able to generate the results you can observe above. The portfolio uses 20 different currency pairs and 20 different genetic programming generated strategies using techniques which can be reliable simulated only on EOD data as they only rely on closed-candle OHLC values on the daily time frame to make trading decisions (entries and exits). Certainly it is interesting to see that many of the strategies look similar (for example EUR/JPY and GBP/JPY) but some are dramatically different such as the systems on the EUR/USD and EUR/CAD. Overall it is quite impressive to see the power of this technique and how it allows for the creation of extremely robust and well diversified portfolios across a very wide variety of trading instruments.

Certainly portfolios like this are one of my targets for 2011 and hopefully I will start live testing an approach like this soon within Asirikuy. The framework I am using here – from which you’ll learn more in a few days – will be released next week to Asirikuy members although another framework (which is being developed with the help of another great Asirikuy programmer) will also come out this year. The other framework – being C++ coded and independent of MT4-allows for a huge degree of freedom and the ability to code much more elaborate ideas (more about this on a later post).

If you would like to learn more about my work in automated trading and how you too can build your own strategies based on sound trading tactics with very robust simulations please consider joining Asirikuy.com, a website filled with educational videos, trading systems, development and a sound, honest and transparent approach towards automated trading in general . I hope you enjoyed this article ! :o)

Print Friendly, PDF & Email
You can leave a response, or trackback from your own site.

Leave a Reply


Fatal error: Call to undefined function mcsp_html() in /home1/danielfp/public_html/mechanicalforex/wp-content/themes/modern-style/comments.php on line 71