(By Michael Harris)There are many ways of accomplishing automated trading strategy discovery. The idea is simple in principle and I have described it in more detail in my (out-of-print) books since 1999. In a nutshell, historical data series are used to discover robust strategies that fulfill certain performance criteria based on backtesting results. Implementing the idea is not a simple task, however, as a number of popular methods for accomplishing it produce trading systems that are fitted to the data and fade quite fast.
Some of the methods used for automated trading strategy discovery just by virtue of employed methodology generate highly curve-fitted systems and although some of those systems under selection bias may pass an out of sample test they nevertheless fail in actual trading because of their structure and dynamics. The failure is unavoidable regardless of the sophistication of the algorithms used to discover the strategy. The failure comes primarily from the indicators used as building blocks and the method used to combine them, which is essentially an optimization method of some performance objective based on recursive data-snooping.
In contrast, Price Action Lab™ (PAL) is a software program that can be used for automated trading strategy development without some of the problems that render such a process highly dubious. PAL does not employ genetic optimizers, genetic programming or neural networks, methods that are known to produce fitted and optimized trading systems that may survive out of sample testing due to chance alone. Unlike some other programs that claim automatic system generation, Price Action Lab™ does not search for random pattern formations and it does not perform permutations or combinations of rules and indicators. No lagging indicators, no curve-fitting, no optimization. The algorithm used for the search, scan and p-indicator functions was designed carefully to minimize data mining bias and eliminate data snooping bias. PAL searches for patterns with common characteristics that minimize selection bias. In addition, the program provides a function for a sigle out of sample test and cross-validation to eliminate data snooping bias. All patterns discovered by the search function can be tested on unused data with a mouse click.
Price Action Lab is a true white-box. Here are links to six examples of automated trading system development using Price Action Lab:
The first example is about a system for SPY, the second is about a system for QQQ, the third about a system for EC futures, the fourth about a system for YM futures, the fifth about a system for DIA and the sixth about a system for GBPUSD. Anyone interested in purchasing the code for these systems please follow this URL for the proper email address to use. Full system code is available for Tradestation and Amibroker and condition code for several other platforms.
We strongly recommend working with the demo first to become familiar with the program before placing an order. The demo version can be ordered here.