Read this first and realize that if you use GA to combine trading rules the strategy you find must be significant at (t-test of strategy/# of combinations tried), i.e. it will never be significant for all practical purposes. IMO the best way is to come up with an abstract hypothesis and backtest it without GA data-mining. Good luck.
You'll quickly come to the conclusion that out of hundreds of technical indicators, they are largely variations on just a handful of basic premises. Oscillators, moving averages, volatility range, that type of thing. Time frames are the most overlooked aspect of technical analysis IMHO.
I'm trying to do some data mining of trading strategies by using a genetic algorithm to combine different trading rules and would like to get my hand on some more novel technical indicators than the usual bunch (rsi, cci, etc,..).
Any suggestions?
