|
PureBytes Links
Trading Reference Links
|
> -----Message d'origine-----
> De : Terry B. Rhodes [mailto:trhodes3@xxxxxxxxx]
> Envoye : mardi 19 novembre 2002 19:49
> A : omega-list@xxxxxxxxxx
> Objet : Re: Are the Neural Networks and Fuzzy logic of any use ?
>
>
> Pierre Orphelin,
>
> Thank you for your objective, unbiased views on Safix-X and TS.
> And thanks to all Pierre's happy customers whose postings are a
> re-affirmation of the value of his opinions and products.
>
> regards,
>
Yes, thanks!
The message below has nothing to do with (un)satisfied customers and was
reposted here by permission.
Just to see if a more interesting thread could start from here...
> -----Message d'origine-----
> De : Gary Fritz [mailto:fritz@xxxxxxxx]
> Envoye : mardi 19 novembre 2002 19:03
> A : sirtrade@xxxxxxxxxx
> Objet : RE: Are the Neural Networks and Fuzzy logic of any use ?
>
>
> You train on genuine random data, then trade on real data, and
> make 100%/yr on the real data!?
Yes.
>
> I am totally confused how you could extract any useful training
> from random data that could perform so well on real data. If
> it's *really* random, then there is NO useful information in it,
> by definition!
There are information in it that you can extract.
But no predictive information.
The software extracts a complex filtering scheme that is able to detect trends
of a minimal amplitude that exist in any random data serie...and that are
themselves random.
>
> If it behaves that way, why not just sell the random-trained
> indicators as a Holy Grail for all markets? Or at least for all
> futures indexes?
It does not work ( or really poorly) on unseen random data ( where it conforms
to the theory and to the arcsinus law).
It works on market data, mainly those who have a correct signal to noise ratio,
what is the case for intraday index futures.
It's not the holy grail, simply a proof of two things:
-Markets are not random ( or *less* random than the learning data that are
perfectly random).
-Safir-X is a tool able to easily extract hidden complex filter rules from most
data available.
However I do not say that you must use true random data to train a system.
Just make the same experience with any NNet package and tell me what you are
able to find within 5 minutes.
>
> (BTW I assume these 100% results were on something that's
> actually tradeable, like SP or ND, not the index itself, SPX or
> NDX? Or if you used the index, you added in appropriate
> "slippage" ?)
I of course used futures index , not indexes where the results are far much
better ( signal to noise ratio is better for indexes than for futures).
The test has been done using CAC, FTSE, DAX and EUREX futures.
I dit it last year during a conference and used ND and SP too with similar
results
The experience has been done publicly, it was not prepared in any manner.
>
> > This apply to most of the NN package where you cannot expect stable
> > results before days.
> > This do not apply to Safir-X that is able to find valid solution
> > within minutes and no learning curve.
>
> I'd disagree with the "no learning curve." I downloaded Safir-X
> a year or two ago, poked at it for a while (not very thoroughly I
> will admit) and I was still confused how to run the thing. I
> eventually got busy with other issues and never went back to it.
This is because of my weak english.
The software is really easy to use.
There is a slide show here:
http://www.sirtrade.com/safirmode.htm
>
> I'm sure it looks easy to you. You've been thinking that way for
> years now. But it was not at all clear to me, and I'm not
> stupid.
In never said that.
You are one of the most proficient contributor to this list and I always
appreciated your posts.
P.
================
As I feel that your reaction is of interest, may I CC it with your permission to
the Omega List ?
This will change from my usual VK ranting...
|