|
PureBytes Links
Trading Reference Links
|
-----Message d'origine-----
De : Tom Cathey <K1JJ@xxxxxxxxxxxxx>
À : Omega List <omega-list@xxxxxxxxxx>
Date : dimanche 11 avril 1999 02:48
Objet : Update on the NeuralFuzzy Project
>Well, well. I'm afraid it's come to this.
>
>Just to give you all an update on the Fuzzy project I've been working on
>here.
>
>To refresh memories, I picked up a $2300 NeuralFuzzy package and intended
to
>put my five homespun S&P indicators into it as a trading assistant to help
>take the mental strain off.
>
>Well.... after about 125 hours of serious work the best I can do is get a
>very retarded trading assistant working. It seems that there is so much
>noise in the normal market data and indicators that the network is easily
>confused.
>
>I am beginning to see that the human mind is by far the best network for
>recognizing patterns. It is quite easy to integrate the whole picture and
>discount the noise with the brain, but the software programming is really a
>cumbersome and endless rulemaking process that is sure to be inferior.
>
>I also tried the other normal neural net optimization stuff...all the MA,
>RSI, etc. horseshit.... and it's no better than TS opto routines. I see
why
>I gave up on this in '94, Tudor gave up on it as well as others. I just
had
>to try again.
>
>It's such a relief now to trade with my old method and real time indicators
>instead of babying and second guessing this thing. But I kinda knew this
>would be the result anyway, didn't I ? Learned a great deal and now
have
>links to Excel, HashNumbs, etc. as a result.
>
>Just 'cuz I decided it's not for me doesn't mean these things can't work
>well in other areas of science. I just think the brain is a much better
tool
>when there is so much market data noise to filter out.
>
>Well, no loss. The company has agreed to take it back for a full refund. I
>may be looking to roll the $ into some other worthless software like an
>Elliot counter, etc...who knows? Always looking for new sucker software.
><g>
>
>Later,
>
>Tom Cathey
>
>
What you bought is a general purpose neurofuzzy logic tool.
You have spent 125 hours for nothing.
The time to spend to build a true neurofuzzy trading assistant is for more
than 125 hours.
We have rather spent more than 125 weeks to build our tool WITH NEUROFUZZY
LOGIC SPECIALISTS, ready for use , and I guess that you will not need 125
hours to find something interesting with it.
The cost is nothing because we provide a full working evaluation version.
Just focus on indicators and let it work for you.
You just have made the error that I made several times years ago in
believing that a general AI package will do the trick.
Unfortunately, this is not the case, and building a specific AI application
for trading is not a piece of cake.
I guess that you were impressed by the price of the software ( $4000,ready
for TS with our, $ 2300 and all the work for you with the package you
bought).
The evaluation version is still available from our web.
If you spend 1/10 of the time with it, I should be surprised that the
conclusions could be the same.
Keep this experience in memory when trying to spare $1600.
Sometimes it is not the best choice.
Sincerely,
Pierre Orphelin.
web: http://www.sirtrade.com
|