|
PureBytes Links
Trading Reference Links
|
Well, well, well.
Pierre gave a hint in one of his messages that that
what his software was doing was tweaking the RL, so I
went <Beyond> that.
Without his software, just comparing 3 systems, one
David proposed and 2 of my own, with 2 different RL
settings for my 2 systems and comparing 7,3,1 and
7,0,0 on David's ADV/DECL system(like a 6 unit matrix
overall), choosing the best TNP at end of each month
to trade for the next month, I got TNP of $126,266.00
for all of 1999 with less than a $11,000 MIDD and 9 of
the 12 months profitable.
Never had two consecutive loosing months, and OddBall
(or derivations) was not chosen any of the 12 months
based upon best TNP of the 6 systems of any prior
month.
It cannot be curve fitting because the requirement was
that you had to trade the system that gave the best
TNP in the next future month, regardless. Only at the
end of that month could you reevaluate and choose
another. The theory is based upon the assumption that
there must be a "sweet spot" in market behavior where
the length of lookback time is most valid or most
profitable for predicting performance for given time
length in the future.
Now I need to run it on 2001, plus use a different
twist: As the months progress, choose only the most
profitable system historically overall for that year.
Also, lookback times and trade forward times could be
changed to 2 weeks, 2 months, etc.
Chao
|