|
PureBytes Links
Trading Reference Links
|
Guy,
Yes, that does bias the channel to the upside. The result is that I
get out sooner with more profit when a trend is broken. Unfortunately, the
cost is that I also get out a lot of times when I shouldn't <G>.
Thanks for your detailed explanation and chart. I've printed out
your response and I'll experiment with it on stocks in my database.
Jim
----------
From: Guy Gordon <gordon@xxxxxxxxxxx>
To: metastock-list@xxxxxxxxxxxxx; JimGinVA@xxxxxxx
Subject: Re: Quantum, QNTM
Date: Wednesday, July 30, 1997 11:40 PM
Jim, you wrote:
"I start with the low just previous to the last bounce off the
bottom of the longer term channel and end with the highest high. "
Doesn't that bias your channel to have an up slope?
You are including low data on the left side by starting at a low,
and high data on the right side by ending at a high.
A least-squares fit through that data is biased to point to the upper
right.
Instead, I try to end my Raff line where the middle line lies on a
price bar (rather than a high or a low.)
Second, it should be parallel to trendlines drawn between highs or
between lows.
Third, it should be parallel to a Moving Average. I haven't defined
exactly which MA, but in part that should be determined by the length
of the Raff channel. A short Raff channel should be parallel to a
short period MA. The MA should be as short as possible. It needs to
be long enough to smooth out unwanted peaks and valleys. But too long
a period MA doesn't tell you anything. If you only have one month (20
trading days) of a new trend, a 26-day MA won't tell you anything.
Let's try this for a definition: A Raff channel covering N days
should be parallel to a N/2 day or smaller Simple Moving Average.
Also, you should ignore the first N/2 days of the MA, as it includes
data from before the trend change.
Fourth, on short Raff channels, it is best to have an equal number of
peaks and valleys. In other words, don't include start AND end at
both highs or both lows.
I can't always make a Raff channel fit all these criteria. But the
more the better. If I had to choose among them, I'd say being
parallel to the right Moving Average is most important. Of course,
that assumes the MA is a straight line (else, you don't have a Trend.)
Take a look at the QNTM chart I am posting at
http://www.white-crane.com/qntm.htm. I have a near-horizontal Raff
channel (dark blue) from mid Feb through the first week of July.
Price movement in this period is very sloppy, and I had no luck
drawing a trendline. But the Raff channel is instructive. Note that
I have it start and end where the middle line lies on a price bar.
The 26-day MA (dotted red line) wavers a lot, but is sort of parallel
to the Raff. The 50-day MA (solid red line) is parallel to the Raff
after about 1 1/2 months.
The second Raff (light blue) starts on the turning point at 5/30.
While this is a price low, it falls on the middle line of the new
Raff. This second Raff extends through today (7/30) where the middle
line again falls on a price bar. Note that it is parallel to the
26-day MA, once the MA has a chance to turn up in the second week of
July (right where it crosses the middle of the OLD Raff channel.)
Finally, the second Raff channel is parallel to a trendline
(resistance) drawn through the highs of 6/17, 7/16, and 7/24. In
fact, the top line of the Raff IS that resistance line. (That's nice,
but it is only required to be parallel, not coincident.)
Another thing I do with Raff channels is to play with one end and see
how sensitive the channel slope is to the end position. (I just drag
the endpoint back and forth and see how the channel tilts.) If the
Raff is not too sensitive to endpoint placement I take this to be a
positive, meaning the channel has more validity. If the Raff is very
sensitive to endpoint placement, then I am more skeptical of it.
The reason for this is that a Raff channel is a least-squares fit to
ALL the data covered. So it should not depend, excessively, on the
last point or two.
One more thing that helps validate Raff channels is that the price
bars sometimes cluster along the middle line. (I wish I knew why.)
Sorry to have gone on so long on this. In fact, if I write any more,
I'll have to give it a bad name, like GordonMath, and charge for it!
Gak! Aaa! Cough, Cough. Excuse me. I nearly choked to death there.
On Wed, 30 Jul 1997 09:34:19 -0400, Jim wrote:
>Guy,
> I see where you drew your channel. What rules do you use for drawing
>raff regression channels. Are you just looking for ones that are almost
>parallel to moving averages? I'm trying to develop hard and fast rules.
>Right now, for short term up trend channels, I start with the low just
>previous to the last bounce off the bottom of the longer term channel and
>end with the highest high.
>Jim
>
>----------
>From: Guy Gordon <gordon@xxxxxxxxxxx>
>To: metastock-list@xxxxxxxxxxxxx
>Subject: Re: Quantum, QNTM
>Date: Tuesday, July 29, 1997 8:58 PM
>
>
>OK, I drew your channel, and QNTM broke it today. And I can see why
>you chose 6/26 as a starting point, as a reversal day.
>
>But I like this channel better: Raff Regression from 5/30 to 8/25.
>Note that 5/30 is ALSO a reversal day.
>
>The reason I like this channel better is that the lower line
>corresponds to the 26-day MA for almost the entire month of July.
>And the top line of the Raff corresponds to a line draw through the
>highs of 6/17, 7/16, 7/23, and 7/24. These two things validate the
>trend for me.
>
>
>----------
>
>
>
----------
|