Over the past year I have experimented with three different methods of trading in the currency markets: computer automated algorithmic trading, human manual trading, and semi-automated algorithmic trading.
Purely automated trading – using pattern recognition and data mining – has proven to be very unsuccessful, which is is to be expected, as it is highly unlikely that there exists a magic formula for technical analysis that can be applied at all times. If such a thing existed, the designer of the magic money box would quickly become the richest person in history, given the trillions of dollars that pass through the financial markets every day. Successful automated algorithms in the markets compete on speed of response to price imperfections, and to variances between market expectations and economic data announcements.
Manual trading has had variable returns, yielding a slight net gain, but this is no guarantee of future performance due to the lack of a reusable fully codified strategy.
The most successful method is the hybrid combination of a human and a machine working together in real-time, which has performed very well over the testing time period. The human (me) innovates and modifies ideas from large volumes of data processed by the algorithm, as directed by me. The semi-automated algorithm then dispassionately executes the trading plan in the lifespan of the strategy, instantly and precisely processing every market tick, every moment of every day.