I've spent 1000's of hours manually backtesting dozens & dozens of methods. Of the few that showed a small edge none ever worked real time. I'm trading a method I developed live right now on the NQ. I manually backtested 9 months from July through March. It made money every month except February when it showed a small loss. I'm on my third week & have lost money almost every day? There was never that much of a drawdown during my testing. Guess I'll never get it?
i play with options on es futures. My backtesting was verry simple, going back 7 years, find the begin/end/high/low price between the 2 consective expiration dates month by month, then find how many months (out of 84) has a movement greater than 100 pts by expiration, 95 pts? 105 pts? etc..draw a curve. And how many has movement greater than 100 pts etc.. in between the month as high/low (for margin requirement). Then based on that execute a strategy of writing iron condor / strangle. dunno if it works or not yet, only been doing it for 2 month, so far so good.
I use no more than three calendar months of 1 min bar data for my various cycle projections based on neural network analysis, as well as various Tree classifier data mining methods. I agree with other posters. Anything longer than 3 months is too long. I've successfully trained on no more than the previous 1 - 2 weeks of 1 min bar data as well.
In for more info. Are you still using this system? Did you figure out what was going on? I'm manually backtesting stuff right now and it takes forever. I'm also looking for some good software (that is reasonably affordable) to backtest index futures with 1 min and volume candlestick data. Any suggestions? Thanks.
I think its tough because you really have to pay attention to how your system performs in different volatilities. I had a system that kicked ass for the ER2 in 2006 and early 2007, then died when vol picked up in July and August of 2007. For all those guys that backtest only a few weeks or months, how do you know to switch systems? Do you look at vol, or wait for a huge drawdown. Nothing (in my humble career) has worked 100% of the time, but there are times when some systems work and then others do. Figuring out when to switch off and on is the real trick. I have one system that I currently run that has been profitable for energies for the past 14 months (using 3 min chart). I have put it up on this site, but never get any responses... This is one of the best threads in a while. roy