|
PureBytes Links
Trading Reference Links
|
Bilo:
Often your posts set me off in a whole new direction, but this one had the effect of
confirming the approach I've already embarked on. This is especially timely because
I've just made the decision to write my own software to implement a system approach
that includes most of the elements you mention, along with some related ideas.
For example I have some ideas about how to represent Timothy Morge's "trade
location" concept as a fuzzy logic rule. I.E., instead of just "buy now because I
think the price will go up", it's "put in a limit order at the bottom of a coil that
will most likely resolve to the upside - this gives a high risk-reward because I can
know quickly if I'm wrong."
This programming project is going to take months of part-time work, so it's very
nice to get some outside confirmation at the beginning that the direction makes
sense. Thanks.
David
p.s. I did find your ai.zip post, and as usual your viewpoints make a lot of sense
to me.
> -----Original Message-----
> From: Bilo Selhi [mailto:citadel@xxxxxxxxxxxx]
> Sent: Friday, June 08, 2001 8:46 PM
> To: David Rosenthal; Omega List
> Subject: Re: Fuzzy Logic
>
>
> one example of fuzzy variable is trend strength...
> very strong, strong, normal, weak, very weak
> another example is volatility
> very high, high, normal, low, very low
> another example is liquidity
> very high, high, normal, low, very low
> another example is trend persistence...
> another one is overbot oversold levels
> very overbot, overbot, neutral, oversold, very oversold
> etc, etc...
> linguistic vars.
>
> those are all fuzzy variables.
> the right approach would be:
> - to know the right variables to use, not letting the machine
> select those for you but knowing which ones to use from
> your knowledge domain ( market rule base )
> - to know the right rule base to use for those variable to
> generate buy and sell signals, not letting the machine find
> the rule but to define those rules yourself...
>
> the FL engine would then look at the combination of
> several key variables and fire off an event or a signal
> or a series of events that lead to a signal
> based on those variable combinations.
> FL allows you to loosen up on the vars and rule
> instead of being rigid on those ( FL vs crisp logic )
> read the basics...
>
> - know the key variables
> - know the key rules
> the results should then match the specs...
>
> instead of
> - let the machine curve fit the variables
> - let the machine curve fit the rules
> so that the results match specs... ( ultimate curvfit )
>
> you are the knowledge domain engineer and YOU
> define the vars and the rules...
> that's why i wanted a clean simple FL toolbox...
> not some software that does that all for me without
> allowing me to do what i want.
>
> trade size selection, risk, reward can be considered
> fuzzy vars also, so no big new revelation there...
>
> which vars and rules i ain't gonna tell you before
> i see the toolbox.
>
> i prefer rule based approach, FL rule based
> approach has advantage over crisp rule based
> approach. that why FL is better.
> i wrote a long post about it some time
> ago. you get in trouble when you go into the NN
> black box based approach where you lose sight
> of your rules....and when that happens you can't
> follow the rules... because you can't see the rules
> or the rule make no sense or the rules are curvefit
> and when you don't follow the
> rules because you don't know the rules then
> you get "ruled" over...
> however market does not like rigid rules
> nor does it like no rules or rules that make no sense,
> ( not entirely deterministic and not entirely chaos )
> so it's somewhere in between where the truth
> lies. simple FL can help out... especially in decision
> making.
>
> bilo.
>
>
>
>
> ----- Original Message -----
> From: "David Rosenthal" <davidrnews@xxxxxxxxxxxxx>
> To: "Bilo Selhi" <biloselhi@xxxxxxxxxxx>; "Omega List"
> <omega-list@xxxxxxxxxx>
> Sent: Friday, June 08, 2001 5:37 PM
> Subject: Fuzzy Logic
>
>
> Bilo:
>
> You said: "on top of that rule based system can be improved by 50-100% if
> utilizing
> even simple FL with the right approach."
>
> Another one of your intriguing statements! This has led me to start
> educating
> myself on the wonderful world of fuzzy logic. Care to expand on what you
> consider
> "the right approach" ??
>
> Thanks,
>
> David R
>
> > -----Original Message-----
> > From: Bilo Selhi [mailto:biloselhi@xxxxxxxxxxx]
> > Sent: Tuesday, May 29, 2001 7:06 PM
> > To: Omega List
> > Subject: Re: ??? developing or purchasing a system
> >
> >
> > i tried my best.
> > what i tried to relate to the list is that it
> > not the technique or the complexity of it that matters
> > it's the approach or the concept that does.
> > a bad concept ( generalizing here no offence )
> > can not be saved by a
> > complex technique and a sound concept
> > can be supplemented well by it.
> > fuzzy logic is just a simple technique that
> > can supplement a good concept if you have it.
> > there is no need to get overly sophisticated at
> > the technique if one does not have sound concepts :-)
> > that's all i have to say.
> > bilo.
> > ps. i would be happy to check out the new safir
> > although my heart tells me that i would rather
> > have a generic FL toolbox for Tradestation :-)
> >
> > ----- Original Message -----
> > From: "pierre.orphelin" <pierre.orphelin@xxxxxxxxxxxxxx>
> > To: "Omega List" <omega-list@xxxxxxxxxx>
> > Sent: Tuesday, May 29, 2001 9:03 PM
> > Subject: RE: ??? developing or purchasing a system
> >
> >
> > > I may understand your position, but the way we go is the opposite.
> > > We do not believe that such a simple tool will be of practical interest
> > > theoretical: yes) fuzzy logic approach needs some fine tuning, what we
> do
> > > in the Safir-X software.
> > > Coding a FL toolbox for TS is going back 5 years for us, and more, we
> use
> > > neurofuzzy logic, not fuzzy logic alone, for various reasons , among
> them
> > > generalization and stability.
> > >
> > > Adding a FL toolbox to TS will bring more complexity to system
> > development,
> > > if you want to mix this with pure EL code.
> > > I do not want to speak of the development and trial and error testing
> > time,
> > > we speak days or weeks, not minutes, like it is the case with the
> > > standalone application.
> > >
> > > If I had to demonstrate the validity of the approach, I would certainly
> > not
> > > propose such a tool, because our experience showed ( with the early
> > version,
> > > 5 years ago) that this kind of approach was time consuming and
> > discouraging
> > > for the end user.
> > > Rather post realtime result, what I will do next month.
> > >
> > > The way we go is to the other side , with even more complex software
> that
> > > will automate the full process of not only setting buy sell neurofuzzy
> > > rules, but also decide if you use stop or limit orders, add or decrease
> > the
> > > number of contracts, go flat, all of this being done with neurofuzzy
> > rules,
> > > 100 % automated research and multi criterion training, all of this with
> > > portfolio capabilities.
> > >
> > > This will be the next Safir-X Professional version scheduled for Q4
> 2001,
> > > targeted to institutional ( price will be too).
> > >
> > > Then you may understand why we do not have any interest to code a FL DLL
> > for
> > > TS, a matter of time, of intellectual interest( that I would have had 6
> > > years ago), and frankly the neat impression that we are doing what
> should
> > be
> > > done in the neurofuzzy approach for trading systems development, even if
> > we
> > > are not the best of the best ( we are not).
> > >
> > > Thisis the reason why we cannot do this now, and do not want to support
> a
> > > technique that is difficult to use without the help of a specialized
> > > software, what a TS DLL is not.
> > >
> > > Sincerely,
> > >
> > > Pierre Orphelin
> > > www.sirtrade.com
> > > TradeStation Technologies representative in France
> > >
> > >
> > >
> > >
> > > > -----Message d'origine-----
> > > > De : Bilo Selhi [mailto:biloselhi@xxxxxxxxxxx]
> > > > Envoyé : mardi 29 mai 2001 21:38
> > > > À : pierre.orphelin@xxxxxxxxxxxxxx; Omega List
> > > > Objet : Re: ??? developing or purchasing a system
> > > >
> > > >
> > > > actually Pierre,
> > > > the best thing you can probably do for all of us on the list
> > > > since you are the expert in FL field
> > > > is to code a generic FL dll implemented toolbox.
> > > > it has to be purely generic ie. simplified functions
> > > > such as normalize, fuzzify, infer, defuzzify and that's about it...
> > > > this would greatly improve the rule based only TS approach...
> > > > why not do that and capitalize on that instead of getting
> > > > into real complex stuff that only a handful will buy???
> > > > you can pick up the Earl Cox book ( i can mail you one if you want )
> > > > at amazon it has most of the code you need on the CD in there.
> > > > simplify it, compile it into a DLL and dump it on the market
> > > > for $300 bucks or something like that... many will thank you
> > > > for that i am sure.
> > > > on top of that rule based system can be improved by 50-100%
> > > > if utilizing even simple FL with the right approach.
> > > > if you get the dll done i can demonstrate the right approach to
> > > > you and to the list...
> > > > i hope you understand.
> > > > bilo.
> > > >
> > > > ----- Original Message -----
> > > > From: "pierre.orphelin" <pierre.orphelin@xxxxxxxxxxxxxx>
> > > > To: "Omega List" <omega-list@xxxxxxxxxx>
> > > > Sent: Tuesday, May 29, 2001 2:06 PM
> > > > Subject: RE: ??? developing or purchasing a system
> > > >
> > > >
> > > > > Could you explain ?
> > > > >
> > > > > Pierre Orphelin
> > > > > www.sirtrade.com
> > > > > TradeStation Technologies representative in France
> > > > >
> > > > >
> > > > > > -----Message d'origine-----
> > > > > > De : Mark Brown [mailto:markbrown@xxxxxxxxxxxxx]
> > > > > > Envoyé : mardi 29 mai 2001 18:53
> > > > > > À : Omega List
> > > > > > Objet : ??? developing or purchasing a system
> > > > > >
> > > > > >
> > > > > >
> > > > > > ps i already warned them that anyone who was a vendor replying
> > in
> > > > > <============?????
> > > > > > private was most likely a scammer who wanted to get into
> > their
> > > > > > pockets. so if your such a legitimate vendor, you won't mind
> > answering
> > > > > > and or responding in public i would think? of course watch
> > out
> > > > > > for anything fuzzy. be nice now i am..
> > > > > >
> > > > > >
> > > > > > Have Great Day, Mark Brown
> > > > > > orgies organized
> > > > > > rebellions quailed
> > > > > > bs 2000 support
> > > > > > fuzzy stuff vacuumed <=================???????????
> > > > > > wedding cakes
> > > > > > planets aligned
> > > > > > anything else to make a buck, almost.
> > > > > >
> > > > > >
> > > > > >
> > > > >
> > > > >
> > > >
> > > >
> > >
> > >
> >
> >
>
>
|