Hi, I am a relatively new Metastock user and I have put together some system tests that seem to backtest OK but I am running into some unexpected results when I try to tweak the system. THe system is as follows: Buy: C > (Trough(1 ,Mov(C ,5,S ) , 2) + (C * .01)) AND Ref(C,-1) < (Trough(1 ,Mov(C ,5,S ) , 2) + (C * .01)) Sell C < (Peak(1 ,Mov(C ,5,S ) , 2) - (C * .01)) AND Ref(C,-1) > (Peak(1 ,Mov(C ,5,S ) , 2) - (C * .01)) Now I would like to improve the system by only executing the trade if the Price cross the H (for buys) after the day of the signal or crosses the low for sells. So I tried coding it like this: First I created custom formulas so the Buy is now Fml( "Trade Buy") and Sell is Fml( "Trade Sell") Buy: X:=ValueWhen(1, Fml( "Trade Buy") ,H); C>X Sell: X:=ValueWhen(1, Fml( "Trade Sell") ,L); C<X Now I would expect better trades with this tweak but the opposite occurs, ie the equity curves get much worse. I am incline to think that the reason is my coding and not the usling H and L crosses as a filter.

Beags Jose is right. Systems built around ZigZag-based functions cannot be becktested reliably with a system tester. My in-depth testing from some years ago (using other methods) found that one particular Peak/Trough-based system produced between 5% and 45% of phantom signals, depending on the non-ZigZag filters I applied to it. Typically, a signal appearing on the current bar will disappear two or three bars later as new data is added. All you actually test with the Enhanced System Tester are the signals that remain valid with subsequent data. Phantom signals will not feature in your results but will get you into real trades if you act on system signals in real time. Roy MetaStock Tips & Tools