The Trend Machine

There is, however, exciting and promising research using nonlinear methods and modeling techniques culled from the sciences of complexity and artificial life (A-life). The underlying hypothesis is this: While the basic input data of the markets—primarily prices—may be simple, the output can only be forecast with nonlinear methods derived from complexity theory. They do not use back-fit data or curve-fitting, as do all conventional technical analysis methods. These include chaos theory, catastrophe theory, and cellular automata (CA). CA essentially grows a forecast from a seed using an algorithm or set of algorithms. Simple CA algorithms can generate complex behavior—just as the basic buy and sell orders lead to the great variety of chart formations. Whether it is possible to beat the markets with them remains to be seen. For an example, see “A Simple Cellular Automata Model for Predicting FX Prices,” by Michael Duane Archer, given to the Automata 2008 conference in the United Kingdom. A link to it is available on www.goodmanworks.com. I suppose one could call this digital trading. Do not laugh; I coined the phrase algorithmic trading in 1996, and am still awaiting royalties.

Figure below shows a 12-hour noninterpreted forecast for the EUR/USD in 1-hour increments from the trend machine. A 1 is a forecast for an up bar; a 0 is a forecast for a down bar.

Trend Machine Forecast

The forecast can be run in a stacked semaphore, with high, low, and close rather than just close. It can also make forecasts interpreted to ME directional movement from –4 to +4, where –4 represents a steep downtrend and +4 a steep uptrend.

A recent back-test on the EUR/ USD with 15-minute bars going back to January 2005 showed promise. Of 16 ME states, the program performed extraordinarily well in 9, profitable in 5 others, and poorly only in 2.

Considerable research is being conducted on a wide variety of nonlinear models; someone will make a breakthrough in the not-too-distant future.