|
PureBytes Links
Trading Reference Links
|
Well, we have a couple of clients putting millions of real ZAR into
this simple model, in real trades, making real money....
OK, in all honesty, this model is part of a huge array of models that
we use, for real clients, trading real money etc etc, but it is a
significant part of that array of models.
Apologies for the theory, but a lot of recent research has made it
clear that one should use a suite of models, not just one. We have
devised all sorts of pratical ways to, in a theoretically sound
manner, combine the results of many different models into something
that leads to sound decision making in the real world, but does not
shy away from the theory or the intricate implementation required to
make it work in practise.
But don't just look at us.... Many central banks nowadays utilise a
suite of models rather than just a single model as they did just a
couple of years ago. A lot of the jargon comes from central banks'
modelling teams. The theoretical and practical reasons for this are
many. Some models work better in the short term, some in the long
term, some are better for cyclical markets, some for trends. Some
give you target levels, some models provide extreme values. The
theta model that I am sharing with this group is an amazingly simple
model that proved itself (don't believe me, read the article) in a
variety of applications and for a variety of time series.
I think it has relevance to the real world of trading and was
requested by members of this real world group to provide code for it.
This I did, for your viewing pleasure and perusal - a right, which I
note, you are exercising to the full.
Regards
MG Ferreira
TsaTsa EOD Programmer and trading model builder
http://tsatsaeod.ferra4models.com
http://www.ferra4models.com
--- In equismetastock@xxxxxxxxxxxxxxx, "Jose" <josesilva22@xxxx> wrote:
>
> MG,
>
> > One where the exponential decay weight is optimised for.
>
> Optimized to what, equity curve, market cycles? Curve fitting, maybe?
>
>
> > Of course it works. In a scientific study, referred to before,
>
> The real question is, does it work in the *real world*, with *real
> trades*, using *real $*. Or is it just another paper study/theory,
> dressed in quant jargon, with little or no relevance to the real world
> of trading?
>
>
> jose '-)
>
>
> --- In equismetastock@xxxxxxxxxxxxxxx, "MG Ferreira" <quant@xxxx>
> wrote:
> >
> >
> > Comments below,
> >
> > Regards
> > MG Ferreira
> > TsaTsa EOD Programmer and trading model builder
> > http://tsatsaeod.ferra4models.com
> > http://www.ferra4models.com
> >
> > --- In equismetastock@xxxxxxxxxxxxxxx, "Jose" <josesilva22@xxxx>
> wrote:
> > >
> > > Please excuse my ignorance, but curiosity knocks:
> > >
> > >
> > > "...but I think that will not apply to most series studied in
> > > Metastock."
> > >
> > > Why not?
> >
> > I think most people use Metastock for end-of-day or financial market
> > data, where there usually is not a clear seasonal trend. I know
> > there are exceptions, but 'usually'....
> >
> > > And what is an "optimal exponential moving average"?
> > >
> >
> > One where the exponential decay weight is optimised for. This is
> > what many people will understand as an exponential moving average or
> > exponential smoothing. Technical analysis guys do not optimise for
> > this weight, they specify it either by giving the length parameter
> > or, if the software allows it, explicitly giving the weight.
> >
> > >
> > > "The long term component can be interpreted as the long term or
> true
> > > value of the market, say where the market is supposed to be based
> on
> > > fundamentals."
> > >
> > > All this from a Time Series Forecast indicator... What kind of
> > > fundamentals are you referring to?
> > >
> >
> > Note the 'say where the market is supposed....' Of course, no real
> > fundamentals are used, but it makes it nice and practical to
> > interpret it that way. Note the tsf function is used as a proxy,
> > the real theta model is slightly more sophisticated. I think it is
> > not too difficult to implement it correctly in Metastock actually,
> > at least the long term portion, but we often use this same approach
> > in practise for a variety of reasons.
> >
> > >
> > > Is this (Theta) theoretical stuff, or does it apply to trading the
> > > real markets? If so, does it work?
> > >
> >
> > Not entirely sure what you mean with 'theoretical' vs 'real markets'
> > and 'does it work'? Are these supposed to be exclusive concepts?
> >
> > But, acting on what I suspect you have in mind.....
> >
> > Of course it works. In a scientific study, refered to before, this
> > model was found to be one of the best forecasting models when
> > compared to many others. 3003 series from all walks of life,
> > financial, economic, demographic, bussiness etc. was tested, in
> > many different frequencies (daily, monthly etc.) and the Theta model
> > really fared well. It was pitched along side every conceivable
> > exponential moving aveage technique (figuratively speaking), neural
> > networks, RBF models and others and outperformed the lot. It is all
> > in that article that I referred to before if you need to check the
> > salient details.
> >
> > Now there is a big difference between forecasting and trading, but,
> > if you apply the Theta model correctly, it will work in trading too.
> > The code I posted does not implement it as theory dictates, but is a
> > practical enough implementation for Metastock to handle it, yet
> still
> > in the spirit of the default Theta model (theta = 0, 2).
> >
> > We usually use both the Theta lines and the Theta projection itself,
> > as well as an extrapolation of the latter, and had excellent results
> > with it. Using the three lines referred to is a lot like using
> three
> > indicators at the same time - it gives lots of additional
> information
> > but adds a bit of subjectivity as well, which is often nice in a
> real
> > trading environment.
> >
> > Finally, since I claim that, if used correctly, it will work, please
> > note that the usual disclaimer applies ;-) I cannot nor do I
> > guarantee that it will work in any circumstances and can assure you
> > that it sometimes won't work, even in the most ideal of
> circumstances,
> > but it sure is worth a try.
> >
> > >
> > > jose '-)
> > > http://www.metastocktools.com
> > >
> > >
> > >
> > > --- In equismetastock@xxxxxxxxxxxxxxx, "MG Ferreira" <quant@xxxx>
> > > wrote:
> > > >
> > > >
> > > > Hi there,
> > > >
> > > > Some people have asked about the Theta model. I have not used
> the
> > > > Theta in Metastock before, but made a quick and dirty
> implementation
> > > > of it which you will find below. The Theta model decompose the
> > > series
> > > > into two components, a long and a short term one. If you use
> > > monthly
> > > > data, then it has a third component, which is the seasonality,
> but I
> > > > think that will not apply to most series studied in Metastock.
> In
> > > the
> > > > default Theta model implementation, you have theta = 0 (long
> term)
> > > and
> > > > theta = 2 (short term) and you fit an optimal exponential moving
> > > average
> > > > to the short term, but to do this in Metastock will be quite
> > > difficult.
> > > >
> > > > The long term component can be interpreted as the long term or
> true
> > > > value of the market, say where the market is supposed to be
> based on
> > > > fundamentals. The short term component is where the market
> finds
> > > > itself based on shorter term factors. Thus the shorter term
> factors
> > > > pushing it away from the long term line. The projection itself
> is
> > > > exactly midway between the long and short term lines.
> > > >
> > > > Anyhow, we normally use both the short and the long term lines
> in
> > > > strategies, as well as extrapolate the theta line itself, which
> is
> > > > very easy to do (but difficult in Metastock).
> > > >
> > > > Regards
> > > > MG Ferreira
> > > > TsaTsa EOD Programmer and trading model builder
> > > > http://tsatsaeod.ferra4models.com
> > > > http://www.ferra4models.com
> > > >
> > > > ------------------- Theta model Metastock code follows
> > > >
> > > > {MG's implementation of the Theta model
> > > >
> > > > This is not an exact implementation,
> > > > it e.g. does not optimise the theta = 2
> > > > weight, assumes a theta = 2 line (as is
> > > > usually assumed, reducing the whole model
> > > > to a linear fit and an exponential fit to
> > > > the residuals for the theta = 0 and
> > > > theta = 2 lines) and the theta = 0 is not
> > > > really a straight line, but the last value
> > > > of length amount of straight lines fitted
> > > > to the data. But the spirit of the model
> > > > is the same as that of the real theta.
> > > >
> > > > This is actually a forecasting model, thus
> > > > the theta line is a projection of where the
> > > > price is supposed to go to. To use in
> > > > practise, take it as a type of moving average.
> > > > Thus if the price cross it, buy as if the price
> > > > just cross above a moving average and vice
> > > > versa.
> > > >
> > > > MG Ferreira
> > > > 2005
> > > > http://www.ferra4models.com
> > > > http://tsatsaeod.ferra4models.com
> > > > }
> > > >
> > > > {Length to use when fitting}
> > > > length:=Input("Length",5,1000,50);
> > > >
> > > > {Theta long term line, assume theta=0, sort of...}
> > > > lt:=TSF(CLOSE,length);
> > > >
> > > > {Theta short term, assume theta=2, sort of...}
> > > > rs:=CLOSE-lt;
> > > > st:=lt+2*Mov(rs,length,E);
> > > >
> > > > {Theta}
> > > > (lt+st)/2
------------------------ Yahoo! Groups Sponsor --------------------~-->
In low income neighborhoods, 84% do not own computers.
At Network for Good, help bridge the Digital Divide!
http://us.click.yahoo.com/EpW3eD/3MnJAA/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/
|