If you want to know whether this system has an edge in the market I would do the following; - Build a watchlist of 50 stocks for each sector; consumer staples, consumer discretionary, industrial, energy, healthcare & technology. - Use the same rules to test each sector...eg..20% of the low & 20% of the high.....same position sizing aswell. - Test each watchlist independently using the following timeframes (1995 - 2000), (2000 - 2003) & (2003 - 2006). Doing this gives you an good overview of how the system will perform under different market conditions as those above periods encompass high volatilty, low volatilty, bull markets, bear markets, price spikes & sideways trends. Even after doing the above you still may not have a true reading of this system's robustness because of survior ship basis....the only way to overcome this is start backtesting with the current S&P500/NASDAQ100 index as of 1995...don't start backtesting with today's index components. After doing this you can expect to find wildly different results for each watchlist and in each timeframe....tech in '98 and '99 would make you a millionaire whilst tech in '00 and '01 would put you back in the doghouse. Hindsight is perfectly clear. Trend-following systems can and do work but only under the right market conditions as when the tide comes in all boats rise. Good luck.
Its a good thing to have in your bag of tricks to put into action after major market lows to catch the next uptrend. But at other times it can be very problematic. I've created trend following systems that could have solved our national debt problem since Oct 02. But run them in the previous two years and you'd be looking for bankruptcy protection. Personally, I don't like to put myself so much at market risk, completely dependent on the market environment.
What you call curve fitting I call using a method that shows a profitable outcome over a long period of time. I do not know why you feel that all securities must show the same price behavior and must be traded with the same trading method. An alternative to using a method that shows a profitable outcome over a long period of time is to use a method that has zero or unprofitable outcome over a long period of time. Nobody knows if ANY system will show profit in the future.