|
PureBytes Links
Trading Reference Links
|
This post brings up an interesting point (to me anyway); one that I've been grappling w/ for awhile. Simply, how does one know (most probably) whether his trading system results are random or not?
Best Regards,
Bill Vedder
"pierre.orphelin" wrote:
> To make short this long story , I started to write on this french group list one week ago.
> The main concern of the yats list was to demonstarte that technical analysis may only work by chance, and that trading systems are no more than unconscious biased overfitting.
>
> So, I proposed a little experience based on a data serie that was showing a random 50% up and down move and a random 50% of a 11% amplitude or 50 % 10 % amplitude.
> The original question posted was how to profit from these theoretical data,only knowing the above hypothesis.
>
> Knowing their real aversion to technical analysis, I said that I was able to solve this with the data provided by them as a Gif screen, showing obvious trends.
> And use for that a well known neurofuzzy product where the training data will be the cac 40 intraday files that are provided with the Safir-X evaluation version.available for free.
>
> They sent the raw data , and the same day received a bunch of neurofuzzy trading systems results that were working fine on the supposed random data, unseen to the system that I trained on cac40 data and sp500 intraday data as previously announced without any modification of the training data
>
> More, then indicator data derived from their original serie were provided to them in order to verify by themselves that I had not overfitted anything ( the original training data being provided with the evaluation version of the Safir-X softare, unchanged since 9 months or so, and they got the unseen data.
>
> So they decided that I was a liar , that it was impossible, that I was stupid and so and so, not speak of usual commercial reproachs ( you know what I mean).
>
> I replied with technical comments and a lot of screen shot showing that their so called distribution was anything you want but not random, the reason why the system worked and why I accepted the test without having the data serie, because it was obvious only by eyeballing their screenshot.
>
> The more the explanation were provided to explain their error , the more they refused to admit that the data were not random, because they were generated by using the Excel pseudo random generator ( they however never provided the worksheet).
> And because I could not succeed with the data because they were randam . Period. And here starts the infinite loop.
>
> Then Mark Brown was there ( probably starting to learn french on yats), copied the french text ( translated below) of the offensed falsely random data author,without any objectivity, as usual.
> Not forgetting to ask for my removal ( the english to french translator that he used is far worse than the reverse version)
>
> The so called random data characteristics provided by them were as follow:
>
> Kurtosis far above 3, skewness >0.15 for various c-c[n] distributions, an obvious 12.5 bars frequency on average
> and Hurst coefficient determined by a contributor to the o list as > 0.5).
>
> Useless to say that the neurofuzzy tool was bound to succeed with such a random data set, trained on anything you want.
>
> At this time, and due to the noise done on this list ( pros and cons PO, not to speak of my own message attempting to calm the waters), they consider to close the list ( I have proposed them to remove me in order to save their lost honour, or what remains).
>
> Comitting suicide because of a trivial error made by building a bad devised trap where I avoided to fall, and not willing to recognize it, this has not been seen in this field, since the old Japan kamikaze era.
>
> I sincerely hope that they will finally recover from their psychological defeat.
>
> Life is a sad thing...
>
> Rgds
>
> PO
> www.sirtrade.com
>
> -----Message d'origine-----
> De : Mark Brown <markbrown@xxxxxxxxxxxxx>
> À : omega-list@xxxxxxxxxx <omega-list@xxxxxxxxxx>
> Date : lundi 18 décembre 2000 15:07
> Objet : The Yats Group
>
> >Have you wondered what PO has been up to? Well he has landed on a
> >French Trading List where they have become tired of him as well.
> >yats@xxxxxxxxxxx
> >
> >Anyway its funny to see a fellow french man call PO arrogant and to
> >ask the moderator of the list to refrain him from Fuzzy promotion of
> >his products.
> >
> >the below was written by one of the french members of that list.
> >
> >Hello ,
> >
> >Orphan Mr pretends that the software that he sells foresaw some things
> >to the seen of the set of the courses of AlexPlanck Inc. while
> >training on the returns of the SP500 or the CAC or me doesn't know
> >what stock indication....
> >
> >It is probable that he/it is a good programmer, but him is probable
> >that he/it didn't understand anything at the financial markets and it
> >is certain that he/it doesn't have any common sense, what is the most
> >serious. It is in all cases a very bad seller.
> >
> >I don't try to demolish Orphan Mr nor to disparage his/her/its
> >software. Seen the way of which he/it entered with arrogance a
> >discussion some course while throwing a challenge of which he/it
> >didn't measure the consequences, him, only had that for that he/it
> >asked. As for the tone that I used, he/it doesn't have been not very
> >different from the one that I sometimes use on the list and who
> >irritates some of it. But Orphan Mr follows this list since one
> >moment, and while applying to me he/it had an idea of the way of which
> >me would answer.
> >
> >I decided myself therefore one-sidedly auto to censor
> >
> >To finish, last advice,:
> >
> >1 - since the object of the list is not either masochism, nor the
> >treatment of the obsessional delirium, frustrations or the mental
> >confusion, I suggest to Orphan Mr to be going to see elsewhere for to
> >make itself/themselves hit over, since he/it seems to like the
> >strokes, or to take care of his/her/its neuroses, since he/it seems to
> >make pathological fixings on the YATS list and his/her/its members.
> >
> >2 - to the moderator, I recommend to follow the suggestion of D Namur
> >and to censor this thread, or even to curb Orphan Mr seriously him
> >even. I call those that share this point of view (or the inverse) to
> >himself to make know (by private promenade of preference, it is
> >cluttered enough like that here; -) to help the moderator to decide.
> >In that that me concern, me him formless that work is made: I am auto
> >- moderate to to leave of now; -)
> >
> >To soon on other topics,
> >Grumpy also; finally! (it is rare!)
> >Alex.
> >
> >
> >
> >
> >--
> >
> >Best regards,
> > Mark Brown mailto:markbrown@xxxxxxxxxxxxx
> > Y = Offset + Amplitude * sin(Frequency * X)
> > .o©º°¨¨°º©[ WWW.MARKBROWN.COM ]©º°¨¨°º©o.
> >
> >
> >
> >
|