|
PureBytes Links
Trading Reference Links
|
Hi,
You are right when using a few classical systems that will have
unpredictable equity curve behavior because they only represent an
approximation that often works more by chance than anything else ( the
case is general for any false science, what technical analysis is).
This is less true with more sophisticated adaptive system (
neurofuzzy logic base for example) that have thousands of rules and
that may better adapt to unknown market condition). They will fail
some day and encounter some drawdown, but probably later that
classical hand coded EL systems.
This is where the difference resides , and the slightly better
stability of the equity curve of neurofuzzy systems, when allowed to
compete together, allow to get a valid results when swapping the
systems, the small winning equity curves are detected sooner ( or will
last longer) and have a better chance when aggregated to produce the
expected result ( we attempt to minimize the lag for swapping detection
, increase the duration of detected winning trades sequence in this
case, otherwise what you said will occur if you have to pay twice
the lag of the equity curves added to the lag of the underlying system
themselves (what is the drawback of the equity curve trading).
It's NOT guessing since we have realized this ( check the web site
below)..
It has been verified and tested over years on multiple supports (
mainly index futures and main forex pairs - range bars in this case-
).
The stock picking version will be ready for September, coupled with a
TS scanner to select the stocks.
Moreover this new approach is also a good answer to the usually
annoying questions :
-When should I stop the system(s)
- What should I do now ?
... Because the systems are permanently swapping and the losing
systems are quickly discarded ( the 1000 candidates are evaluated in
real time), so there is always something to do with the latest best
sorted candidates.
Useless to says that this works with a managed portfolio of several
lines (even small ones, but 10 in a minimum IMHO, although even 1 or 2
produce better results than a single sytem.
Sincerely,
Pierre Orphelin
www.sirtrade.com
16th year of automation in trading system development
-----Message d'origine-----
De : DH [mailto:catapult@xxxxxxxxxxxxxxxxxx]
Envoyé : jeudi 20 août 2009 04:10
À : Omega List
Objet : [english 100%] Re: Switch between systems
I'm with Gary. Dynamically switching between systems is too little too
late. By the time you see the system failing and switch away, you've
already lost too much. An alternative is to do size ~ (sharpe-ratio)^2
among the systems.
--
Dennis
|