Hello everyone, I'm looking for some experienced trading system developers to give me a hint: After some serious drawbacks in my trading system design...including postdictive errors, curve fitting, you name it ...I've decided to start building my road to success all over again. Lately I've been consuming a lot of the Van Tharp materials...where the role of expectancy plays a central role in trading system development...which as far as I can judge is a plausible concept. How do you go forth in developing a positive expectancy system? For eg.: - "I define my entry and exit strategy...then backtest with a fixed amount position sizing...if the system has a positive expectancy, I start tweaking the position sizing" I find it difficult to split up the process (position sizing, entry, exit), since when calculating the expectency of a system the topics are interwoven with each other. All insights, recomendations, books to read, etc. are welcome.
I recommend you obtain high-end backtesting software and begin to study the relationship between "Expectancy" and "Overall System Goodness" as defined by You. You may find, as I do, that Expectancy is a very poor predictor of how well you are going to like a system's results. In fact you may conclude that Expectancy is misleading (!). Attached below are backtests of the same trading system, with two different parameter settings: the "Blue" parameter setting, and the "Red" parameter setting. The system uses a Breakout entry: buy on a Stop at the Highest High of the past N bars. It also uses a Breakout exit. It's traded on a diversified portfolio of more than 70 futures markets around the globe. I ran the backtest simulator with a number of parameter settings, and chose to display these two. The Blue setting has a low Expectancy but a high "desirability to me". The Red setting has a high Expectancy but a low "desirability to me". Not only are the returns of the Blue setting higher (CAGR = Compound Annual Growth Rate, in percent per year); so are the gain-to-pain ratios such as Sharpe and MAR. The measures of "pain" (longest drawdown in months; maximum drawdown in percent) are worse for the high-Expectancy Red setting. By "high-end backtesting software" I mean something that will trade a portfolio of >100 instruments simultaneously, using fixed-fractional and dynamic positionsizing. Something that will also allow you to trade several different systems simultaneously, each with its own unique portfolio, out of the same account. I know that Trading Blox Builder, Mechanica, Trading Recipes, and PowerST can do this. I have heard it said that AmiBroker and Wealth-Lab can do it too, but haven't seen it with my own eyes.
Hello MGJ, thank you for your very interesting reply! ...unfortunately the topic doesn't seem to interest a lot of people... I've been thinking about the results you've posted. What was especially intriguing for me is, that the systems performance results are differing with a factor of 20! The only explanation that came into my mind is: The opportunitys taken by the higher performing system are more...which basically means that it does more trades. Am I right? Van Tharp calls this "Expectunity" (a mix of expectancy and opportunity). Heres an example: Trader 1: Is a trend follower who does one trade every 20 day. His trades have an expectancy of 2.38. This gives him an opportunity factor of 0.05 (since he does 0.05 trades per day) Trader 2: Is a day trader generating 500 trades a day with an expectancy of 0.11. Expectunity is now opportunity times expectancy: Trader1: 0.119 Trader2: 55 Therefore although Trader 2 has a lower expectancy rate, when factoring in opportunity, it becomes clear that he will make a lot more money a year than Trader 1 since he can harvest a lot more opportunitys Does this equally apply to your two systems?
Might want to look into the liquidity of that expectancy too. That way you get a system that scales well.
I wonder if you are analyzing and researching to avoid success. Possible trying to justify changing your system. Perhaps you are trying to avoid volatility and seek certainty. I give up seeking certainty. I accept, even embrace volatility. My trading system does not mean that I will be profitable next year. My system shows that limiting risk and placing small open ended bets occasionally shows a profit. My system is more of a general guide to keeping more of what I want. It does not mean that I won't experience losses.
How do you define expectancy? You may find the free article by Michael Harris "Derivation of the Profitability Rule and its Application in the Discovery of Trading Systems Based on Price Patterns" very interesting. Also read the next free article "Automatic Discovery of Trading Systems: The Next Step in Mechanical Trading" to find out how you can have a computer find a system for you. Here is the link: http://www.tradingpatterns.com/About_Us/articles/articles.html I recommend Metastock for starters and Wealth-Lab or Tradestation for advanced system developers. John
John, take a look at a new article by Michael Harris on expectancy: http://www.tradingpatterns.com/About_Us/articles/articles.html Ron
MGJ, I agree with you when you say that expectancy is not everything. However, your data demonstrate another important point to me : a system's metrics (characteristics) may change in time. This is clearly the case with your two sytems : I rescaled them at different points in time. The differences between the two curves tend to desepear arround the 3000 - 3100 th trade. At that moment, your curves still have 6 years to go, which is a lot (1500 trades still to be taken). Hence, not only is it important to look at several metrics. It is essential too, to check them time by time. Olivier.