To: David R. Evans who wrote (10311 ) 5/26/1999 2:16:00 PM From: Alex Stangl Read Replies (2) | Respond to of 12039
Thanks for the warm welcome, everyone. Sorry I haven't started contributing anything here yet. I took Richard's advice and read through (and captured significant posts from) the TA Indicators & Systems thread. I've entered a bunch of systems in to MetaStock and tried benchmarking them on various stocks from the IBD Weekend Review list. I find myself craving more capabilities in the MetaStock system tester. It would be nice if the System Tester could more closely emulate the real-world performance of the system, and also properly reflect that XX% profit made in a short-period is much better than XX% profit made over a longer period. (I tried to somewhat account for the latter by making the interest rate 20% or so, to make it give more emphasis on systems that remain invested for shorter periods.) Anyhow, one of systems that seem to be providing the best overall, consistent returns so far in my testing seems to be VHF (with varying periods & threshold -- but 27/0.3 seems to work well.) One thing that bothers me about VHF, though, is that it appears to not distinguish uptrends from downtrends. But every time I try to add something in to the formula to try to make it only generate buy signals on uptrends, it appears to worsen the performance of the system! I'm sure many of you have run across the same sort of thing, but it's a little paradoxical, and frustrating to boot. I'm suspecting that stocks on the IBD list have experienced overall uptrends in the past, so that the VHF is appearing good even without distinguishing downtrends -- but it could be a killer going forwards, when the inevitable downtrends do start to occur. So I'm thinking maybe of doing system tests over a wider range of stocks. But it's frustrating that for all the work that the computer automates, that it doesn't do more. I've been tossing around the idea of writing some system testing software to interface with the QP2 DLL and be able to go through the entire database (or a subset), trying out different systems across different stocks, across varying time periods, etc. I'd be able to put in proper stop loss signals and have it more properly account for time invested, and money management. It'd be a fair deal of work, though. Even more so, if I don't hardwire stuff, and try to make it generic and usable by others. Do you think this'd be a worthwhile endeavor, or would I be better advised to continue learning, and maybe invest in some other software capable of doing more extensive system testing? Best Regards, Alex