[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Tr: MB ads revisited (I)



PureBytes Links

Trading Reference Links

 ----- Message d'origine ----- 
De : <tradejacker@xxxxxxxxxxxx>
À : <omega-list@xxxxxxxxxx>
Envoyé : jeudi 2 décembre 1999 21:35
Objet : Re: MB ads revisited


> and to trade a lousy stochastic...what a f***king deal at $5000 a pop!!!!
> 
> more snake oil revisited ala fuzzy pete.....save your money, folks, and 
> learn this trading shit for yourselves, there are no easy outs. and starve 
> a frenchman to boot :))
> 

Yes, I have had some stupid reflections like this one on an other list and decided to perform
 quick test with only ONE indicator: Wilder's RSI (14) .
Could have been stochastics or anything you want.

Here are the results found by Safir-X at this time

Maybe you will finally understand that indicator works with PROPER RULES.
These rules are not discloses in any book.
Only a software is able to discover them.

I wait your counter example where you are able, with a single indicator like RSI,
to find winning systems on unseen data and without optimization on these data.
Results below are given for daily bars.
Of course, intraday data produce far more better results ( 2x to  3x  ROMID factor ).


1) Daily data used  
DJI (1977==>1999)
FTSE
SPX 
CAC
DAX
(8-10 years)
 
2) Indicator used for training in SafirX :

 RSI and previous values ( first differences)

This is what you have to code in TradeStation:
==================================

Value1=rsci(c,14);
Value2= ddf(value1,1);  ( means rsi- rsi [1])
Value3= ddf(value1,10);  
Value4= ddf(value1,30);  

3) 
Training data
 DJI  from 1977 to 1983  5 years only.
 
 
4) Test on unseen data ( out of sample) :
FTSE
SPX 
CAC
DAX
 .
5 ) Results :

Header Legend ( left to right)
======================
One line= one system
One column= one data set ( train or test data)
 
S=# solution ( means system) 
Partial view due to omega List  size limitation
 
                   ROMID % results : see below 
Chk4= DAX
Chk3=CAC
Chk2=SPX
Chk1= FTSE
Train= training on  DJI (1977-1983)
Test=  test on DJI (1984-1999)
Whole= train+test on DJI (77-99)
Nb Ru...= nb of fuzzy rules  per system
NbFS= nb of fuzzy sets per system
sorted by decreasing performance on DAX ( Could be any thing you want)
 
Development time:
< 3 min ( train and test on unseen data)
Yes you read 3 min, and  100 solution were found, and tested
 
Enjoy!

 Sincerely,

-Pierre Orphelin
Neurofuzzy Logic tools for TradeStation
Free evaluation versions and competitive upgrades available
web: http://www.sirtrade.com