Okay, I suppose I could actually try to answer your questions instead of
just plugging my package =)
1) Instead of random trades in the real data, the typical test is to
repeatedly construct random data and see what the strategy would do
2) I think this distribution would be the same as all k-period returns
(assuming 5 minutes between trade n and trade n+k). I'm not sure this would
be helpful
3) Since you have to compare your signal's performance against something,
comparing to 0 is one approach. Another is to compare to being long all
day.
Good luck!
-David
[[alternative HTML version deleted]]
_______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-sig-finance-- Subscriber-posting only.
-- If you want to post, subscribe first.
