SPY2p5 Trading System Forward Performance

The SPY2p5 system was machined designed by Price Action Lab using an in-sample from SPY inception (01/29/1993) to December 31, 2008. Then the system was validated on unseen data, called the the out-of-sample,  from January 2, 2009 to April 5, 2012 and the results were published in this blog two days later. By the end of last year the SPY2p5 system had outperformed the ETF. This post presents detailed forward results of the system from April 5, 2012 to the close of last Friday.

The steps for using Price Action Lab to design the SPY2p5 systems were presented in the April 7, 2012 post. The same results will always be generated by the program when using the same in-sample data.  Price Action Lab is based on a proprietary algorithm that produces the same output each time it encounters the same conditions and this determinism is in compliance with the standards of scientific testing and analysis.

Below are the in-sample results obtained using unadjusted SPY data from Yahoo. These results can be confirmed even with the demo version of the program (the demo will calculate performance results but will not generate code for implementation in other platforms):

SPY2p5res

In-sample performance

Below is the in-sample performance of the SPY2p5 system of patterns for the in-sample period, first if multiple positions by the various patterns are allowed and then when no multiple position are allowed. Version 4.o was used to generate the results. This version will be released by the end of this month:

SPY2p5_IN

It may be seen that if multiple positions are allowed the system generated 459 trades with win rate 72.33% and profit factor equal to 2.67. Next is the performance with no multiple positions, i.e. only one long or short position is opened at a time and new similar signals are ignored:

SPY2p5_IN_NMP

In this case 316 trades were generated with win rate 69.30% and profit factor equal to 2.34. The results did not change a lot when no multiple positions were allowed.

Out-of-sample performance

The out-of-sample performance with multiple positions allowed is shown below:

SPY2p5_OOS

In this case the system generated 99 trades with win rate 64.69% and profit factor equal to 1.93. The performance is satisfactory. Next are the results with no multiple position allowed:

SPY2p5_OOS_NMP

With no multiple position allowed the system performance degrades due to a large drawdown but it is still acceptable with win rate 55.71% on a sample of 70 trades and profit factor equal to 1.30. Note that this drawdown regards only the drop in the net profit of the system and not the drop of bankroll. The latter will depend on risk percent and starting account size. The percent drawdown when positions are sized so that the account is fully invested and for starting capital of $100K turns out to be about -12%. What appears as a large drawdown on the equity curve above will be much less and in the order of -2.5% if, for example, the risk percent allocated to this system is in the order of 20% of bankroll in the case of the $100K starting capital.

Forward performance

The main question is not so much how the system performed on unseen but nevertheless past data, the out-of-sample, because this is actually the past but how it will perform on a forward sample, i.e. on future data that did not exist at the time of development and thus data snooping of any kind was not possible. Below is the performance of the system from the end of the out-of-sample period to the close of last Friday with multiple position allowed:

SPY2p5_FW

The SPY2p5 system generated 28 trades with win rate 67.86% and profit factor equal to 2.13. The net profit for this period was 35.98 points. Given that the SPY rose 18.7 points in this period, the SPY2p5 system with multiple positions outperformed the underline ETF by a factor of 1.9, or its performance was about double. Next are the results with no multiple position allowed:

SPY2p5_FW_NMP

In this case the system generated 17 trades with win rate of 70.59% and profit factor equal to 2.40. Note that the SPY2p5 system generated 24.78 points per share in this period and thus outperformed the market by a factor equal to 1.32, or its performance was better by 30%. In the case that reverse position are let to cause an exit, the performance increases to a profit factor of 3.80 and Sharpe ratio of 3.51, when using Amibroker (Charts created with AmiBroker – advanced charting and technical analysis software. http://www.amibroker.com/”) to backtest the system of patterns combined using the OR Boolean operation, separately for longs and shorts. This performance is really outstanding for a machine designed trading system but future results may be different.

Also note that the profit factor in the forward sample is slightly better than that of the in-sample for no multiple positions.

Question: Will the SPY2p5 system stellar performance continue into the future and for how long?

Answer:  Nothing is certain with the markets and trading and no such predictions can be made. The system may continue working for several years or degrade immediately from now on. It is very important when using systems to have some rules for discarding them. These rules can vary widely based on objectives. One advantage in using Price Action Lab is that the system can be updated every year, for example, for the purpose of incorporating in the design new conditions, by moving the in-sample forward by one year. This is what I call Forward System Development (FSD).

Question: Can I find similar systems in other markets?

Answer: Possibly but keep in mind that it is the market that decides what is possible, not you or the software you are using. Successful system developers let the market and the software guide them to what is feasible. Usually, trying to get what you want from the market and the software is not possible. This is quite often a fundamental mistake of many system developers. For example, they struggle with 1-minute data and try to get the software, any software that designs trading systems, to fulfill their objectives. By doing so they usually end up with a curve-fitted system developed based on data-snooping bias. The system usually fails immediately. Then they often blame the software and the market instead of their methodology.

Question: Are the patterns in the SPY2p5 system specific to a single market? How has the SPY2p5 system performed when use to trade other ETFs?

Answer: See the Post “A Robust Trading System Across the ETF Board”. The forward performance has been exceptional even in the case of non-correlated ETFs

Disclosure: no relevant position at the time of this post.

Disclaimer

Related posts

SPY2p5 system original post:
http://www.priceactionlab.com/Blog/2012/04/machine-generated-spy-patterns-and-trading-systems/

SPY2p5 system forward performance for 2012:
http://www.priceactionlab.com/Blog/2013/01/the-price-action-lab-spy2p5-system-outperformed-the-market-in-2012/

This entry was posted in Price Action Strategies, Trading Strategies and tagged , , . Bookmark the permalink.