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

Re: [EquisMetaStock Group] Absolute reference



PureBytes Links

Trading Reference Links

Thank you Eric!

Now, herewith some additional ideas.  One, of course, is to add more
indicators.  Another is to start using combinations of indicators. 
This is the neural network approach.  You could create a new one, say
T4, which is something like this:

T4 := (T1 + T2)/2; {Simply the average of T1 and T2}

Since T1 and T2 are standardised, this is also standardised.  Or you
could create one that measures if both T1 and T2 are positive,
something like

T4 := ( T1 > 0 ) * ( T2 > 0 ) - ( T1 < 0 ) * ( T2 < 0 );
  {+1 if both are positive and -1 if both are negative}

Just do an image search on google on 'neural network' and you'll
quickly see lots of images giving the structure of a typical neural
network.  Note how many 'indicators' are combined into less indicators
and this is finally combined into just one indicator - which is more
or less what we are doing here.  We use the Kalman filter (more or
less) to grow or shrink dynamically the weight we assign to an
indicator (a *real* neural network will optimise a weight and use the
single weight - we don't do that here, and this approach has its
benefits, and drawbacks....).

Anyhow, good luck with furhter work on this.

MG Ferreira
TsaTsa EOD Programmer and trading model builder
http://www.ferra4models.com
http://fun.ferra4models.com 


--- In equismetastock@xxxxxxxxxxxxxxx, chichungchoi <no_reply@xxxx> wrote:
> Hi MG Ferreira:
>   Thank everyone, and specially thank to MG, who makes a very 
> valuable contribution on Kalman Filter. I will keeping exploring more 
> on this managing subject.
> Thank you very much
> Eric :>
> 
> 
> --- In equismetastock@xxxxxxxxxxxxxxx, mgf_za_1999 <no_reply@xxxx> 
> wrote:
> > Thanks TecloGeo and Eric,
> > 
> > This is half of what I am looking for, although the
> > 
> > > sum(C*(Cum(1)=1))
> > 
> > example I included in the post does the same.  Since an array, such 
> as
> > C, can have practically unlimited values, this could be used as a 
> way
> > to bridge the MSFL 20 variable limit.  You could do something like
> > 
> > Variables := 0;      {Unlimited number of variables, all = 0}
> > 
> > Now, to get say 50 variables in a formula, you could do something 
> like
> > 
> > Sum( Variables*(Cum(1)=1) +      {Returns first variable}
> >      Variables*(Cum(1)=2) +      {Returns second variable}
> >      Variables*(Cum(1)=3) +      {Returns third variable}
> >      ...
> >      Variables*(Cum(1)=50) )     {Returns 50th variable}
> > 
> > The problem with this part of the solution is the complexity.  It
> > would have been a lot easier if we could just write say
> > 
> > Sum( Variables[1] +
> >      Variables[2] +
> >      Variables[3] +
> >      ...
> >      Variables[50] );
> > 
> > But we can at least 'read' the variables.
> > 
> > The problem is that we need to also assign values to these - we need
> > to be able to 'write' to these variables.  This is the missing 
> half. 
> > So if we could do something like say
> > 
> > Variables[1] := 20;
> > Variables[2] := Variables[3]+0.5*Variables[4];
> > Variables[3] := xyz;
> > 
> > then we have a really neat way in which to keep all the say weights
> > together, manipulate them, and even plot them.  But I don't think 
> this
> > is possible in MSFL.
> > 
> > Anyhow, Eric, I think using MSFL the best is to first stick to only
> > one PC and one MS.  To optimise each one you would do something like
> > what you propose, although that assumes that the various indicators
> > are completely independant of one another.  This is sometimes 
> assumed
> > to simplify things, but is bound to give suboptimal answers.
> > 
> > You could ensure that each indicator has a minimum of say 10% in the
> > final function, via something like
> > 
> > TT := ( (W1+0.1)*T1 + (W2+0.1)*T2 + (W3+0.1)*T3 ) /
> > (W1+0.1+W2+0.1+W3+0.1);
> > 
> > or you can just give say W3 a very slowly adjusting smooth function,
> > which calls for individual smooth functions!
> > 
> > Regards
> > MG Ferreira
> > TsaTsa EOD Programmer and trading model builder
> > http://www.ferra4models.com
> > http://fun.ferra4models.com 
> > 
> > 
> > 
> > 
> > 
> > 
> > 
> > --- In equismetastock@xxxxxxxxxxxxxxx, "TecloGeo" <teclogeo@xxxx> 
> wrote:
> > > MG,
> > > 
> > >  
> > > 
> > > Is this what you're looking for?
> > > 
> > >  
> > > 
> > > Valuewhen(1,cum(1)=1,C)
> > > 
> > >  
> > > 
> > >   _____  
> > > 
> > > From: equismetastock@xxxxxxxxxxxxxxx
> > [mailto:equismetastock@xxxxxxxxxxxxxxx]
> > > On Behalf Of mgf_za_1999
> > > Sent: Monday, May 16, 2005 7:49 AM
> > > To: equismetastock@xxxxxxxxxxxxxxx
> > > Subject: [EquisMetaStock Group] Absolute reference
> > > 
> > >  
> > > 
> > > Hi Eric,
> > > 
> > > I am working on your question, just a quick question to the group,
> > > that relates to our discussion.
> > > 
> > > Is it possible to get an absolute reference?  I know
> > > 
> > > Ref(C,-1)
> > > 
> > > gives the close one bar ago.  Also, something like
> > > 
> > > sum(C*(Cum(1)=1))
> > > 
> > > gives the close of the first bar.  But is there a function, 
> say 'X',
> > > so that
> > > 
> > > X(C,1)
> > > 
> > > gives the close of the first bar?
> > > 
> > > Thanks in advance
> > > MG Ferreira
> > > TsaTsa EOD Programmer and trading model builder
> > > http://www.ferra4models.com
> > > http://fun.ferra4models.com 
> > > 
> > > 
> > > 
> > > 
> > > 
> > > 
> > >   _____  
> > > 
> > > Yahoo! Groups Links
> > > 
> > > *	To visit your group on the web, go to:
> > > http://groups.yahoo.com/group/equismetastock/
> > >   
> > > *	To unsubscribe from this group, send an email to:
> > > equismetastock-unsubscribe@xxxxxxxxxxxxxxx
> > > <mailto:equismetastock-unsubscribe@xxxxxxxxxxxxxxx?
> subject=Unsubscribe> 
> > >   
> > > *	Your use of Yahoo! Groups is subject to the Yahoo!
> > > <http://docs.yahoo.com/info/terms/>  Terms of Service.




------------------------ Yahoo! Groups Sponsor --------------------~--> 
Has someone you know been affected by illness or disease?
Network for Good is THE place to support health awareness efforts!
http://us.click.yahoo.com/Rcy2bD/UOnJAA/cosFAA/BefplB/TM
--------------------------------------------------------------------~-> 

 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/equismetastock/

<*> To unsubscribe from this group, send an email to:
    equismetastock-unsubscribe@xxxxxxxxxxxxxxx

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/