|
PureBytes Links
Trading Reference Links
|
I too stopped using Raff channels because they are too wide. I've used
parallel lines which are closer to the highs and lows of the data. Any
linear regression line and channel will not necessarily divide the data
equally above and below the mid linear regression line. The linear
regression line is the one that minmizes the sum of the squares of the
points from the mean. This is not identical to an average line.
The way that I use these chanels is to start them at a high or low, and end
them before the last data point(s), I then extend the channel to the right.
In this way I can see if the more recent data is within the channel of the
older data.
I have not noticed the dynamic qualities of the Raff regression channel.
Have I missed something, or does the way that I have used them in the past
prevent this dynamic change?
Lionel Issen
-----Original Message-----
From: J.W.E. Roberts <jan.roberts@xxxxxxxxxxxxxxxxx>
To: metastock@xxxxxxxxxxxxx <metastock@xxxxxxxxxxxxx>
Date: Monday, October 26, 1998 1:40 PM
Subject: Re: Raff channels
>Hi Anton, Don,
>
>> Attached is a better "as example" giving screenshot of the misleading and
>> most offensive Raff floating trend channels.
>>
>> As you can well see these channels will cost a fortune when
>> taken seriuosly.
>>
>> They ARE RECALCULATED EACH and EVERY DAY and as such are HYPER
>> ACTIVE and dynamical and are basicaly moving averaged double
>
>For quite some time now I've been using regression 'analysis' (RA) to great
advantage, i.e.
>making money for me. I think we should take a step back from the whats &
hows, and have a
>look a the basics. Allow me tell you how I evolved the use of RA. I started
out using the
>Raff Regression Channels as found in MS. As you **can only draw these by
hand** - which is
>why they're call "line studies" in MS, btw, a visual chart-analysis is the
only option,
>which in turn means that no mechanical trading system using them is
possible in MS. This
>however doesn't bother me overmuch insofas as I am a 'concentrator' (on
only a couple of
>issues) instead of a 'scanner' (trying to pick my trading positions from a
mountain of
>candidates). Besides, I **like** pouring over a chart and trying to divine
what it's telling
>me.
>
>The most difficult part of this kind of RA is of course
>A. deciding when to start a new short term channel (STC), and
>B. deciding on the start- and end-points for the regression-line.
>I bought Gil Raff's book, and he isn't all too clear on these issues
himself. It's more of
>an art and nearly none of a science, really. I can only advise too learn by
using it on old
>data in historical charts, then step forward and see if you were right. In
the course of
>time one develops a feeling for it. Lastly it's helpful to draw a zig-zag
which is often
>shows up major (and minor) turning points; I also look at the
Kagi-Representation regularly,
>and I often switch between dayly & weekly representations . All three
methods help to
>separate the chaff from the corn, i.e. noise from real moves.
>
>What started bothering me about Raff Channels is that they're often much
too wide to be of
>use. Take for instance the giant downward spike of October last year that
plays havoc with
>nearly every indicator - Raff Channels as well. So I switched to the
Standard Error Channel,
>the width of which you can vary by using different standard deviations. I
thereby 'pinched'
>the channels enough to ignore meaningless - at least in *my* eyes - spikes.
In doing this
>channels become of course more susceptible to false signals. However, as
this procedere is
>sanctified by the Highest Authority - 'His Holiness' Jack Schwager who
calls them "Internal
>Trendlines" - I don't worry overmuch ;-)
>
>The next issue worrying me was that when using Std.Error Channels as well
as Raff's
>Regression Channels there is an unspoken assumption that deviation from the
regression line
>is alway as big to the upside as it is to the downside. Now, just look at
any chart and
>you'll quickly see that this is patently false. So I started experimenting
with parallel
>lines with different distances from the regression line, and that was and
is the perfect
>solution for me - a truly revolutionary concept... Until I discovered that
my Great
>Invention had been stolen by Mr. Andrews long before I thought of it; with
truly devilish
>cunning - not for nothing did he call it the ***Pitchfork***.
>
>Pun & Fun aside, I did go the way from Raff's Regression Channels via
Standard Error
>Channels to Andrews' Pitchfork and it does work for me. Anton Maas mentions
that using the
>Close for the regression line may not be the right choice. Maybe one could
experiment with
>Regression Lines of the Highs and the Lows; I did a quick experiment with
them but I
>couldn't get any meaningful results, so for the time being I guess I'll
stick with the
>Pitchfork as is. Finally, please don't think that I'm among those 'one
dimensional men' who
>place their whole wellbeing (or financial ruin, more likely) on just one
indicator or way of
>looking at a chart.
>
>Your comments & critique would be heartily welcome. Happy Trading,
>Jan Willem Roberts
>
>
|