hello Richard,
Thanks for continuing to post the Cooper scans.
Recently, I went back to a system test I got from one of your previous
posts
A quickie version of Bill's work...
If((Ref(CCIE(13),-1)<-100) AND (CCIE(13)>=-100),1,0) +
If((Ref(Mov(C,5,T),-1))-(Ref(Mov(C,35,T),-1)<
Mov((Ref(Mov(C,5,T),-1))-(Ref(Mov(C,35,T),-1)),4,T)),1,0) +
If((Mov(C,5,T)-Mov(C,35,T)>Mov(Mov(C,5,T)-Mov(C,35,T),4,E)),1,0)+
If(MDI(3)>Ref(MDI(3),-1),1,0)
buy: fml("bill X derek")>2
sell:ccie(13)<-100
I had plotted this a few months ago, and thought it tested fairly well; when I went back to it this week, however, I noticed that i had
inadvertently written the buy signal as
buy:fml("bill X derek")<2
This was puzzling, since the system was testing OK, so I created another test with the correct buy parameters. I couldn't get the correct formulation to beat the incorrect one, and finally ended up trying the following
buy: fml("bill X derek")<2
sell:fml("bill X derek")>2
To my surprise, the system using these erroneous buy and sell signals seemed to beat the original by a factor of 2:1, tested on 100 stocks. Of the stocks on which the correct buy signal performed better, the incorrect buy signal was often fairly close, with both being in clearly positive territory
Just wondering if you have any comment on this anomaly. currently trying to keep up with the developments re: BNS on the other thread...
Thanks, Nikku |
