|
PureBytes Links
Trading Reference Links
|
David,
I'm pretty sure the test was right since I dumped it from MetaStock.
First of all, the ROC is a test based on intermediate term trends and needs
some time to stabilize. I always test with at least 1000 days of data.
Second, and more important, that test performs well on the smallest
portion of the stocks in my database. The test that performs best on most
stocks is my test 01, next 02, and finally the ROC test 03. By using all
three, I get good results on most stocks. Try the ROC test on C-CUBE Micro
(CUBE) for example.
I'll copy an old post on my investment strategy below. It's important
to note that I don't use the tests for buy and sell decisions, but only to
help identify stocks for my watch list. My final buy and sell decisions
are based mostly on trend channel analysis, although I will tighten stops
if I get a sell signal.
When I add a new stock to my database from my Telescan searches, I'll
do a comparison with several MetaStock system tests to see which one works
best for that stock. I'm currently using three and I'll copy all of them
below. I picked the three tests such that at least one of them usually
gives good results - over 100% profit for the 1000 days tested. If none of
the tests give good results, I'll delete the stock from my database. There
are always a few sports out there <G>. Then I change the indicators on the
top of the chart so I can see by a glance, which test works best for this
particular stock. Anytime, I pick a stock for my watch list, I'll rerun
the comparison to make sure to make sure the test indicated still works the
best.
-----------------------------------------------
Jim,
I use TIP searches all the time. I only keep a small database of
stocks in MetaStock and add to it as a result of Telescan searches. I'll
copy some posts below that describe my strategy and my searches. I've
updated the searches since that post, but am sending it since it explains
my search criteria. Yell if you want the modified searches. I just use
the results to identify stocks I'm interested in and haven't tracked the
results on all stocks identified since the TIP search is only the first
step in my investment methodology.
I have a couple of Telescan Prosearches I use to identify new stocks
for my database. I run those searches weekly and look at the Telescan
charts. If you have Telescan and would like to have my Prosearches, just
yell. If I like the charts, I then download the historical data to my
MetaStock database. When I get the stock in my MetaStock database, I use a
default template to set up a CandleVolume chart of the stock with a small
volume window under it and three indicator windows over it. The indicators
are RSquared, LinRegSlope, and MFI. I then run the five MetaStock system
tests in the compare mode, to find which one works best with that stock.
If the mfi test does best, I don't change anything. If CMO, QStick or CCI
does better, I change the MFI indicator to the one that does best. If ROC
does best, I change the Rsquared and MFI indicators to ADXR and ROC. That
way I can tell which test does best by just looking at the chart. About
once a month I look at all the stocks in my database and prune those that
I'm no longer interested in.
I update the stocks in my database daily using the Downloader. The
stocks in my database are in different subdirectories for LEAP Stocks,
Stocks, LEAP Watch, Stock Watch, and portfolio lists. I run my MetaStock
Exploration which is a normalized version of the system tests weekly. For
the stocks identified by the Exploration, I look at the chart and run the
appropriate system test. If the stock looks like it may be a near term
buy, I'll move it to the appropriate watch list. I look at all the charts
in my watch list daily. Once they move out of a buying range, I move them
back to the appropriate stock or LEAP directory. Whenever, I have buying
power in one of my portfolios, I make my selection from one of the watch
lists.
That's a short summary of my investment strategy. Yell if you have
any questions.
Frank,
A quick summary of my investment strategy. First I only keep a small
database of stocks in MetaStock. I try to keep it to less than 200. I add
stocks weekly based on the results of Telescan searches and prune
periodically.
I run MetaStock system tests weekly to find stocks for my watch list
(I'll copy previous post with my system tests at end of this message). I
draw long and intermediate term trend lines and intermediate and short term
trend channels. Trend lines use Trader Vic methodology. Short term trend
channels are Raff regression channels. When I open a new position, I pick
from watch list. I like to pick stocks that are close to the channel
boundary and have a recent system test buy signal. I set targets at the
opposite channel boundary and stop just outside the channel. If market is
trending strongly, I'll leverage portfolio by using buying power to
purchase some long term, slightly out of the money call options. Have been
moving more and more to LEAPS with over a year to expiration. I close a
position whenever a target or stop is hit. For options, I use same target
and stop methodology, plus I'll roll or close whenever there is four months
left to expiration. This seldom happens as I'm usually out with target or
stop hit, but is added insurance to avoid the accelerated time decay
associated with the last few months of option life.
That's it in a nutshel <G>. System test follow:
All,
I've been playing with the new MetaStock 6.0 profit system tests using
our R2 and linear regression slope tests. The good news is that Rsquared
and LinRegSlope are now built into MetaStock and essentially give the same
results as our R2 and Regression Slope custom formulas. There are also two
new Tushar Chande indicators that I'm using - Qstick and the Chande
Momentum Oscillator (CMO). Qstick provides a way to quantify candlesticks.
It is a moving average between the open and close prices. Chande
developed the CMO to measure pure momentum. It's related to, but unique
from other momentum indicators such as RSI, Stochastic, Rate-of Change,
etc.
I've developed four separate profit system tests based on our previous
work with regression analysis. For each stock in my database, I click on
the new MS 6.0 System Tester and select all four tests. Then I click on
the compare box to check it and click on the compare button. When
MetaStock finishes the comparison, I click on Reports and see which test
does the best for that stock. I then plot the CCI, CMO, MFI, or Qstick as
the third indicator above my stock chart depending on which test does the
best. The first two indicators are Rsquared and LinRegSlope. That way I
can tell by looking at the chart which profit system test works best on
that stock and only run it. I'll repeat the comparison every two or three
months to be sure that the results are the same.
If you have the new MetaStock 6.0, give this system a try and let me
know what you think. Any comments or suggestions will be appreciated. I'd
like to get a dialog started on MetaStock profit system tests. The new
MetaStock 6.0 profit system tests follow:
NOTE: I DIDN'T COPY THE OLD TESTS - CURRENT FORMULAS AND TESTS FOLLOW:
----------------------------------------------
MetaStock for Windows Indicator Builder
ROC C Dema1
2*(2*Mov(ROC(C,55,%),21,E))-Mov(Mov(ROC(C,55,%),21,E),21,E)
S/C (LinRegSlope/Close)
(2*Mov(10000*LinRegSlope(C,34)/C,21,E)) -
Mov(Mov(10000*LinRegSlope(C,34)/
C,21,E),21,E)
------------------------------------------------
MetaStock for Windows System Tester
===========================================================================
01_R2,S/C,MFI (Vol Required)
SIGNAL FORMULAS
---------------
Enter Long:
Alert(RSquared(C,21) < 0.15,13) AND
Fml("S/C") > opt1 AND
HHV(Fml("S/C"),5) = HHV(Fml("S/C"),13) AND
HHV(MFI(55),5) = HHV(MFI(55),13) AND
HHV(TSF(C,233),5) = HHV(TSF(C,233),13)
Close Long:
LLV(TSF(C,233),5) = LLV(TSF(C,233),13) AND
LLV(Mov(C,55,E),5) = LLV(Mov(C,55,E),13) AND
Fml("S/C") < opt1
Enter Short:
Alert(RSquared(C,21) < 0.15,13) AND
Fml("S/C") < opt2 AND
LLV(Fml("S/C"),5) = LLV(Fml("S/C"),13) AND
LLV(MFI(55),5) = LLV(MFI(55),13) AND
LLV(Mov(C,144,E),5) = LLV(Mov(C,144,E),13)
Close Short:
HHV(Mov(C,144,E),5) = HHV(Mov(C,144,E),13)
OPTIMIZATION VARIABLES
----------------------
OPT1: Min = -55.00 Max = -13.00 Step = 21.00
OPT2: Min = -89.00 Max = -21.00 Step = 34.00
STOPS ALL OFF
============================================================================
02_R2, S/C, CCI, TSF - All
SIGNAL FORMULAS
---------------
Enter Long:
Alert(RSquared(C,21) < 0.15,13) AND
Fml("S/C") > opt1 AND
HHV(Fml("S/C"),5) = HHV(Fml("S/C"),13) AND
HHV(CCI(55),5) = HHV(CCI(55),13) AND
CCI(55) > 0 AND
HHV(Mov(C,55,E),5) = HHV(Mov(C,55,E),13)
Close Long:
LLV(TSF(C,55),5) = LLV(TSF(C,55),13) AND
Fml("S/C") < opt1 AND
LLV(CCI(55),5) = LLV(CCI(55),13) AND
CCI(55) < 0 AND
LLV(Fml("S/C"),5) = LLV(Fml("S/C"),13)
Enter Short:
Alert(RSquared(C,21) < 0.15,13) AND
Fml("S/C") < opt2 AND
LLV(Fml("S/C"),5) = LLV(Fml("S/C"),13) AND
LLV(CCI(55),5) = LLV(CCI(55),13) AND
LLV(Mov(C,144,E),5) = LLV(Mov(C,144,E),13)
Close Short:
HHV(Mov(C,144,E),5) = HHV(Mov(C,144,E),13)
OPTIMIZATION VARIABLES
----------------------
OPT1: Min = -55.00 Max = -13.00 Step = 21.00
OPT2: Min = -89.00 Max = -21.00 Step = 34.00
STOPS ALL OFF
============================================================================
03_ROC, ema, ADXR, S/C - All
SIGNAL FORMULAS
---------------
Enter Long:
Alert(Cross(Fml("ROC C"),-5),21) AND
Mov(C,21,E) > Ref(Mov(C,21,E),-13) AND
ADXR(13) > opt1 AND
Fml("S/C") > 0
Close Long:
(Fml("ROC C") < -10 AND
Mov(C,21,E) < Ref(Mov(C,21,E),-3) AND
Fml("S/C") < 0) OR
Alert(Cross(opt2,Fml("ROC C")),21) AND
Mov(C,21,E) < Ref(Mov(C,21,E),5)
Enter Short:
Alert(Cross(0,Fml("ROC C")),13) AND
Mov(C,55,E) < Ref(Mov(C,55,E),-8) AND
ADXR(13) > opt1 AND
Fml("S/C") < -5
Close Short:
(Fml("ROC C") > 0 AND
Mov(C,55,E) > Ref(Mov(C,55,E),-8) AND
Fml("S/C") > 0) OR
Alert(Cross(Fml("ROC C"),-opt2),21) AND
Mov(C,55,E) > Ref(Mov(C,55,E),-8)
OPTIMIZATION VARIABLES
----------------------
OPT1: Min = 21.00 Max = 31.00 Step = 5.00
OPT2: Min = 80.00 Max = 180.00 Step = 50.00
STOPS ALL OFF
---------------------------------------------
Sorry for the long posat <G>.
Jim
----------
> From: David Castley <David.Castley@xxxxxxxxxxxxxx>
> To: jimginva@xxxxxxxxxxxxx
> Subject: re: adxr roc
> Date: Sunday, September 14, 1997 1:40 PM
>
> Jim
> I tried the indicators and system test you kindly posted,but, after
double
> checking that I had got every dot and comma absolutely correct, I found
the
> results to be disasterous, typically, nought out of six tests, with a
loss
> of $400 to $500 in 500 days.
> Can you re-check the details posted and see if there are any errors and
if
> not, do you know why the results are so bad? Thanks
|