Into the Future of FOREX

Although it has lost some luster in this century, application of artificial intelligence (AI) methods has been seen in the FOREX arena. The three primary approaches are: expert systems, neural networks, and genetic algorithms.

I developed an expert system-neural network hybrid in the early 1980s, Jonathan's Wave, and used it successfully in the futures markets for a number of years. I moved on to exploring a cellular automata-based model, the Trend Machine (more on this in the next section).
But the possibility of revamping Jonathan's Wave with modern techniques and computer firepower has rekindled my interest in artificial intelligence. The entire AI approach may have a second wind. I continue to predict a resurgence of efforts by the large institutional traders and have already begun to see interest at that august level.

There is intense disagreement on this subject, but I have concluded AI methods are still primarily linear -- no different in underlying structure than a moving average or relative strength indicator. Past market prices and data are manipulated to make forecasts, and curve-fitting remains the theoretical name of the game. The learning of an expert system is simply complex curve-fitting. You cannot teach an old dog new tricks, and if the system is inherently unsound, it will never learn enough to be successful.

The search for a Holy Grail of trading -- a method that will consistently beat the market -- has been afoot from the inception of the markets themselves. In the mid-1900s, many traders published (usually privately) small volumes with techniques to beat the markets. They typically looked good on paper -- but failed when applied to real-time trading. Most were tested on simplistic market environments (trading markets, trending markets) and failed when the real-time market morphed into a different environment.

I am reminded of the secret system used by a trader I met in Hawaii in the 1980s. He believed that the random spray of ink spots from the dot-matrix news printer was actually hidden buy and sell signals from the floor traders. I do not know if he closed his account when the broker went to a laser printer.

The advent of computer analysis in the 1970s and automated trading in the 1990s encouraged traders to use this new tool to find the trading method over the rainbow. Much of the effort has been directed to using vast batteries of conventional techniques with deep mathematical and statistical twists. It is clear, after 30 years of effort: No linear method is going to beat the market, at least not consistently in all markets. As my late programmer, Jim Bickford, would say, "You can torture the numbers, but you can't make them speak."
By Michael Duane Archer
Michael Duane Archer has been an active futures and FOREX trader for more than 35 years. He has worked in various advisory capacities, notably as a commodity trading advisor, registered SEC investment advisor, and branch manager for Heinold of Hawaii. He currently trades FOREX and futures and is involved in several technical analysis research projects.

Copyrighted 2016. Content published with author's permission.

Posted in ...