|
PureBytes Links
Trading Reference Links
|
Traders,
NOTE: TO FULLY UNDERSTAND THIS POSTING
YOU MUST LOOK AT THE .GIF FILES WHICH
FOLLOW -- COULD NOT PUT ALL TOGETHER
THANKS TO LIST HANDLER 65K LIMIT.
As a result of my work with SwingMachine I developed an indicator
(or function) that attempts to provide a continuous method of estimating
the direction and magnitude of a trend (it can be very short -- a large
value for the smoothing constant or very long -- a small value for the
smoothing constant).
The logic for this function/indicator is similar to Wilder's Parabolic
BUT
the "acceleration constant" adapts and attempts to track the slope of
what would be a SwingMachine trend line.
The test vehicle for the experiment I am going to describe is the OEX
starting in 1988 and running until last Friday. For computation of
profit/loss a value of $100 per OEX point was used (the OEX is
adjusted for recent split) and NO commission. For this experiment
I simply wanted to know what the relative good and bad was.
My function is called LEEX3 and the trading system is very simple.
When the Closing price is above the value of LeeX3 then we buy on
stop at the closing price and sell when the price is below LeeX3. A
second long term LeeX3 is also computed at a constant equal to
1/5 that used as the primary length. To go long we must be above
this and to go short the close must be below it. If the close is above
this function and falls below the primary function then we exit the
long position. Vice versa for shorts.
The following data is the result of an "optimization" run using the
"length" (smoothing constant * 1000) for LeeXLen as the
optimization parameter.
Note that around a length of 10 to 15 the system did not do to bad.
Actually it isgreat! However, at a length of 60 (very fast action) the
system got whipped to death and lost a bit of money.
Test LEEXLEN NetPrft L:NetPrft S:NetPrft ROA MaxDD
%Prft
1 10.00 20215.50 20949.00 -733.50 606.69 -3332.10 46
2 15.00 17441.00 17600.00 -159.00 523.42 -3332.10 56
3 20.00 14432.90 17502.90 -3070.00 235.99 -6116.00 40
4 25.00 7525.40 11426.90 -3901.50 88.71 -8483.00
34
5 30.00 8348.20 11613.20 -3265.00 151.13 -5524.00 33
6 35.00 6945.30 10823.30 -3878.00 157.81 -4401.00 33
7 40.00 10250.80 13709.30 -3458.50 249.17 -4114.00 34
8 45.00 9453.20 14018.70 -4565.50 146.60 -6448.10 37
9 50.00 9219.70 15352.70 -6133.00 144.38 -6385.60 38
10 55.00 4805.70 12004.20 -7198.50 69.37 -6928.10 39
11 60.00 -1054.40 9305.70 -10360.10 -12.46 -8461.10 32
12 65.00 2938.20 9145.20 -6207.00 34.73 -8461.10
33
Test AvgWTrd AvgTrd AvgLTrd #WTrds AvgWBars #LTrds AvgLBars
1 3798.42 1555.04 -367.86 6 242 7 43
2 2070.20 968.94 -407.63 10 146 8 26
3 1985.41 481.10 -521.78 12 102 18 20
4 1680.42 198.04 -572.80 13 78 25 21
5 1509.81 198.77 -456.75 14 69 28 19
6 1302.49 150.98 -406.19 15 56 31 20
7 1354.14 205.02 -386.95 17 51 33 17
8 1175.61 165.85 -423.18 21 44 36 13
9 1135.28 146.34 -462.23 24 39 39 12
10 1002.07 69.65 -529.77 27 35 42 10
11 1071.12 -13.35 -515.41 25 35 54 9
12 1028.59 34.17 -445.90 28 30 58 8
Being a firm believer in the Moon Phenomena I decided to run a
simple test USING THE WORST CASE ABOVE with the
exception that in order to buy the close had to be above the value
of LeeX3 and to sell the close had to be below the value of LeeX3,
AND to buy it had to be a new moon and sell on the full moon and we
exited after 8 days. In the second scenario the above was reversed
and we bought on the Full moon and sold on the new and exited
after 8 days.
The results are shown in the attached 2 .gif files.
>From a loss as depicted above to a profit ONLY BECAUSE WE
USE THE MOON AS A KEY ON WHEN AND IN WHAT
DIRECTION TO TRADE.
Remember, this was a part of a very long BULL run and the LeeX3
function tends to point in the proper direction and so it is not really
surprising that going long on both phases of the moon worked in
this instance.
Which way the moon? ? ? ? ?
When the Moon phenomena was applied to a test run with the
smoothing constant set to 10, the New moon buy did much
better than the Full moon buy. Without the moon effect the
system was in the market 1753 days and made $11.50 per
day. With the moon effect the system was in the market
a total of 1032 days (only 58% of the time without) and
made $18.26 per day with less drawdown. This is the real
convincing evidence that you should include the moon
effect in your system if you haven't.
Next study will be a study by years and by months to
see if FULL/NEW varies as a function of month.
Clyde Lee
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Clyde Lee Chairman/CEO (Home of SwingMacine)
SYTECH Corporation email: <clydelee@xxxxxxx>
7910 Westglen, Suite 105 Work: (713) 783-9540
Houston, TX 77063 Fax: (713) 783-1092
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
|