|
PureBytes Links
Trading Reference Links
|
Folks,
I realize it is rather late in a long weekend to be dumping something
like this on you but for those of you with some interest in my thinking
and implementation of the SwingMachine program this is furnished.
First let me mention two names:
1. Henry Wheeler Chase
2. Arthur A. Merrill
The work of Henry Wheeler Chase is buried in a pile of
papers
somewhere in the files of Curt Hessler and maybe elsewhere but if so I
don't know where it is. At one time Curt and Larry Williams were
partners. During this period they acquired all of Henry Wheeler Chases'
remaining documentation about his method for trading from Mr. Chases'
estate. The method was called the Supply/Demand formula and was a very
complicated method of trading. (Papers are dated 12/48).
Part of the Supply/Demand was a method of classification of swings
using ring points. Unfortunately HWC had not organized things
as well as he might and probably some of the things he had wanted to
say kind of got lost.
Arthur A. Merrill wrote a number of wonderful books on market
behavior. To my mind the most significant Filtered Waves - Basic
Theory really laid the groundwork for me to reorganize the stuff I
had gotten from Curt into what for me was something useful. To
give you an idea of when this really got started, I checked my
book and found that Mr. Merrill signed my copy of this book on
082884 even though the stuff I got from Curt came along a lot
earlier than that.
In a later work, Behavior of Prices on Wall Street,
Mr. Merrill talks about M and W waves and gives a listing of the
32 patterns he identifies. In M & W classifications he uses 5
turning points and consequently ends up with a lot more patterns.
In the original book he provided an appendix with the
classification of waves laid out by Dernell Every. Every defined
18 patterns.
I went through this whole mess analyzing the patterns and
came to the conclusion that the four Turning Point method
provided the most realistic answer -- partly because I could
define a set of patterns that were useful and were limited in
number to ten patterns.
The following is a description of and graphic representation of
the concepts of pattern identification I gathered from the above
noted material.
As an ease to definition of patterns we will establish a couple
of conventions.
1. Turning Point - a price at which on both sides of this point
we find the prices either higher or lower than the turning
point price.
2. BottomTP - a low price on each side of which prices are higher.
BTP
3. TopTP - a high price on each side of which prices are lower.
TTP
4. Pattern - a series of prices which encompass two BottomTP's and
two TopTP's or vice versa.
Generally we will consider a pattern to encompass 3 price ZONES.
These zones have a specific value ONLY FOR THE PATTERN which
we are examining and the value is of NO consequence. For convenience
we will call these zones L/M/H for low, middle, and high. These
zones will define the price points at which lie the TTP'S AND BTP'S.
As a final identification of the patterns we describe, there
will be 10 patterns which will be identified by the numeric values
of zero (0) through nine (0).
In order to examine relationships between the TP's in a pattern
we assign alphabetic designations starting with A and ending with D.
The A point is always the earliest TP in time and the D point
is always the latest point in time.
Since a number of people cannot (or will not) accept .gif files
as a means of presenting pictorial information I have drawn the
zone lines and pattern lines using only alphabetic characters. You
may examine the attached .gif to see examples of all 10 patterns
which we identify in our work.
Note that on charts which we present the PATTERN NUMBER (O...9)
appears at price (TP'S) the equivalent of the letter D in the
drawings below.
I have drawn the patterns in pairs representing similar price
configurations in DOWN (on the left) and UP (on the right) trends.
We can examine the relative position of the D point in a
pattern relative to the A point in the pattern and establish
the trend of prices at that moment for all patterns.
The pattern group 5 to 9 have the first leg of the pattern
going from lower to higher prices. Conversely, the pattern
group 0 to 4 have the first leg of the pattern going from higher
to lower prices.
Zone:
. . .B. . . . . . . . . . . C . . .
H /\ /\
. \/. \ . . . . . . . . . / .\/ . .
M A \ / D
. . . . \ . .D. . . A . / . . . . .
L \ /\ /\ /
. . . . . \/. . . . .\/ . . . . . .
C B
Pattern: #5 #4
Zone:
. . . . .B. . . . . . . .C. . . .
H /\ /\
. . . ./. \ . .D. .A. ./ \ . . .
M / \ /\ /\ / \
. . ./. . . \/. . . \/. . . \ . .
L / C B \
. \/. . . . . . . . . . . . . \/.
A D
#6 #3
Zone:
. . .B. . . . . . . . . . . C . .
H /\ /\
. ./. \ . . . .D. A . . . / .\. .
M / \ /\ /\ / \
\/. . . \ . ./. . .\ . / . . .\/
L A \ / \ / D
. . . . . \/. . . . .\/ . . . . .
C B
Pattern: #7 #2
Zone:
. . . . . . . .D. .A. . . . . . .
H /\ /\
. . .B. . . ./. . . \ . . . .C. .
M /\ / \ /\
. \/. \ . ./. . . . . \ . ./. \/.
L A \ / \ / D
. . . . \/. . . . . . . \/. . . .
C B
Pattern: #8 #1
Zone:
. . . . . . .D. . . A . . . . . .
H /\ /\
. . . .B. ./. . . . .\. . C . . .
M /\ / \ /\
. . ./. \/. . . . . . .\/ .\. . .
L / C B \
. \/. . . . . . . . . . . . .\/ .
A D
Pattern: #9 #0
How does all this apply to the swing machine ? ? ? ? ?
Well the first thing the program does is to examine all
swings which exist on the set of data loaded into the program.
The program assigns a classification number (0,1,...,9)
to each pattern it encounters and constructs an array which
identifies each Turning Point (by date/bar number) and the
pattern number associated with the 3 legs/3 turning points
prior to the current point.
The pattern information is used to graph the pattern
leg lines and identification numbers on the chart that is
being constructed.
Once the chart is constructed, the program again reviews
all pattern identifications in pairs of occurrences. For
each such occurrence which matches the last two pattern
identification numbers on the chart, an entry is made about
the two legs which occur AFTER the given id# pair is found.
This entry consists of the slope (computed based on the then
current closing price of the data in the analyzed pattern)
and duration (in bars) of the two FORWARD legs.
The program then sorts the accumulated data based on
the slope of the first leg and "tosses" the lowest and highest
valued slope entries found. Just a safety check to keep the
system from being unduly influenced by a very large or very
small swing.
As the data is accumulated, for every swing that has an
initial leg with a length of 3 bars or more, the swing is
plotted as a projection from the last TP found on the chart.
An option exists to allow the user to select some type
of "AVERAGE" output to be plotted by the program. You may
select either 1, 2, ..., 5 projections to be made. For a
single projection -- all data is averaged. For a projection
of two (2) estimates, the half of the data with the lowest
slope of the first leg is averaged and projected and then
the half of the data containing the highest slope on the
first leg is averaged and projected. The same concept is
followed on higher number of projections.
If you believe that HISTORY REPEATS ITSELF (as you must
if you use any of the Elliot/Gann/..... time price series
projection methods) then you should accept the SwingMachine
as one of the most significant methods of projection of
time and price points that is available today.
The program can accept data in ASCII, CSI, Computrac,
and TC2000 format. The program does not care if the input
is 1 minute, 1 hour, 1 day, 1 week, 1 month, 1 year, or
any other time period you can imagine.
I trust that this will answer most, if not all, of the
questions which to this point I have rather badly answered.
Sincerely,
Clyde Lee
--
Clyde Lee Chairman/CEO email: <clydelee@xxxxxxx>
SYTECH Corporation Work: (713) 783-9540
7910 Westglen, Suite 105 Fax: (713) 783-1092
Houston, TX 77063
|