|
PureBytes Links
Trading Reference Links
|
You can send me email with attachments of any size ... send to
ftonetti AT optonline DOT net
In the posted version of the AFL there is no real intelligence in the
extrapolation ... It is simply based on an nth ( usually 2nd or 3rd )
order PolyFit of the CMA over it's last n points where n = length of
the CMA * x where x is a user parameter ... For whatver reason x
seems to work best in the range of 0.3 - 0.7 which appears to be on
the low side of what I would have expected. The reason I say that is
because the idea ( Hurst's / Millard's not mine ) behind the length
of the CMA is that it should be roughly half of the cycle you are
looking for. So if for example we have a 100 bar CMA then we would be
looking for a ~200 bar cycle. If x is set at 0.50 ( the mid point of
0.3 and 0.5 ) then the AFL will utilize the last 50 bars ( the CMA
Length of 100 * 0.50 ) to perform a curve fit on and then extrapolate
from there. This would seem ?! to be shorter then what I would have
arbitrarily chosen to use since it is only a quarter of the cycle
we're attemping to find. However, outside this range appears to be
too responsive / unresponsive. There are more intelligent ways of
getting the extrapolation to be more in tune with the data i.e. more
like what a human being with a pencil and a french curve ( jeeze, do
these still exist somewhere ) would draw by hand ... I have played
with a couple of ideas of how to make that happen and they are
clearly an improvement of the unintelligent raw extrapolation but I'm
not ready to post the resulting AFL ...
--- In amibroker@xxxxxxxxxxxxxxx, "Rakesh Sahgal" <rakeshsahgal@xxx>
wrote:
>
> Fred
>
> I tried sending 2 graphics to the group to show the difference
between
> the projections and the actual shape the bands take but could not do
> so due to file size limitations. One is the time when the data point
> is in the projection zone. The other shows when the projection have
> moved further and the data point has CMAs' plotted on it.The
question
> I was asking was - is this difference between the projection and the
> actual bands a constraint that is imposed on us when we draw the
bands
> on a PC or is there some other reason for it?
>
> I know you dont recieve emails from the group, so if it is
convenient
> for you I can send the same to you at your mail address as graphics
> embedded in the email . Even though I have saved the two files as
> graphics from within AB as png files, both are quite bulky at about
> 110kb each. So please let me know if it is convenient for you to
> recieve the files.
>
> Rakesh
>
> On 10/14/06, Fred <ftonetti@xxx> wrote:
> > I don't understand your question ... Are you asking why the
shorter
> > length CMA band extrapolations sometimes extend well past bands of
> > longer length ?
> >
> > --- In amibroker@xxxxxxxxxxxxxxx, "Rakesh Sahgal" <rakeshsahgal@>
> > wrote:
> > >
> > > Hi Fred
> > >
> > > For the entire of yesterday before and market hours I have been
> > alternating
> > > between the book and the bands that are drawn by the afl you
> > posted. The afl
> > > is absolutely consistent in it's projections once the CMA is
> > anchored to
> > > a date. That takes care of one big drawback with the only afl
that
> > was
> > > available. There is one thing that I am unable to comprehend.
When
> > the bands
> > > are plotted on the CMA i.e. they have come out of the projection
> > zone, what
> > > appear as Edge Band Trades were visible as breakouts from the
bands
> > to the
> > > upside/downside when that data point was in the projection
zone. Is
> > this by
> > > design/or suggestive of the need for refinement/reworking the
> > projection
> > > methodology?
> > >
> > >
> > > Rakesh
> > >
> > >
> > >
> > > On 10/14/06, brian.z123 <brian.z123@> wrote:
> > > >
> > > > Thanks Fred,
> > > >
> > > > I appreciate guys like you and Tomasz who take me places I
didn't
> > > > know I was going to.
> > > >
> > > > BrianB2.
> > > >
> > > > --- In amibroker@xxxxxxxxxxxxxxx, "Fred" <ftonetti@> wrote:
> > > > >
> > > > > I will post a version ( To the library ) in due course with
more
> > > > > flexibility and "more intelligent" extrapolation ...
> > > > >
> > > > > --- In amibroker@xxxxxxxxxxxxxxx , "willem1940"
<w.j.a.struyck@>
> > > > > wrote:
> > > > > >
> > > > > > Fred,
> > > > > >
> > > > > > Could you also upload the attached files to the library
or to
> > > > the
> > > > > > files section of Yahoo. The attachements are not filed.
> > > > > > Thanks.
> > > > > > Willem Jan
> > > > > >
> > > > > >
> > > > > > n--- In amibroker@xxxxxxxxxxxxxxx, Fred Tonetti
<ftonetti@>
> > > > wrote:
> > > > > > >
> > > > > > > Here's the Hurst DE AFL I previously mentioned I would
> > > > post ...
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > It INCLUDES PolyFit.AFL, which you can get at the AB
> > library,
> > > > in
> > > > > > order to do
> > > > > > > the extrapolations.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > Functionality
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > - You can plot up to 5 lengths ... This can be increased
> > with
> > > > > the
> > > > > > inclusion
> > > > > > > of one or two statements if desired.
> > > > > > >
> > > > > > > - Bands at the moment are percentage oriented, but it
would
> > be
> > > > > > relatively
> > > > > > > simple to allow them to be based on StDev as well.
> > > > > > >
> > > > > > > - Always works up to the bar YOU select by creating
CMA's (
> > my
> > > > > > version,
> > > > > > > slightly modified for additional smoothness ) up to n/2
> > bars
> > > > (
> > > > > > Where n in
> > > > > > > the CMA Length ) prior and then extrapolating from that
> > point
> > > > > on.
> > > > > > >
> > > > > > > - Extrapolation will always be at least to the bar you
> > > > selected
> > > > > > and can be
> > > > > > > extended beyond that by parameter selection.
> > > > > > >
> > > > > > > - Band color change ( Dark to Light ) and Down / Up
> > triangles
> > > > > show
> > > > > > where
> > > > > > > CMA's end and Extrapolation begins.
> > > > > > >
> > > > > > > - The extrapolation portion of any given band can of
course
> > > > > change
> > > > > > after the
> > > > > > > fact for any given bar but once a CMA for a particular
bar
> > has
> > > > > been
> > > > > > > calculated it is in essence frozen from that point
forward.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > There is then NO forward looking of this AFL ... It does
> > > > however
> > > > > > purposely
> > > > > > > offset arrays so that it can plot beyond the last bar
in the
> > > > > chart.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > See attached macro & detailed views and afl ...
> > > > > > >
> > > > > > > <<...>> <<...>> <<...>>
> > > > > > >
> > > > > >
> > > > >
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > Please note that this group is for discussion between users
only.
> > > >
> > > > To get support from AmiBroker please send an e-mail directly
to
> > > > SUPPORT {at} amibroker.com
> > > >
> > > > For other support material please check also:
> > > > http://www.amibroker.com/support.html
> > > >
> > > >
> > > > Yahoo! Groups Links
> > > >
> > > >
> > > >
> > > >
> > > >
> > >
> >
> >
> >
> >
> >
> > Please note that this group is for discussion between users only.
> >
> > To get support from AmiBroker please send an e-mail directly to
> > SUPPORT {at} amibroker.com
> >
> > For other support material please check also:
> > http://www.amibroker.com/support.html
> >
> >
> > Yahoo! Groups Links
> >
> >
> >
> >
> >
>
Content-Description: "AVG certification"
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.1.409 / Virus Database: 268.13.27/517 - Release Date: 11/3/2006
|