|
PureBytes Links
Trading Reference Links
|
<x-html><!DOCTYPE HTML PUBLIC "-//W3C//DTD W3 HTML//EN">
<HTML>
<HEAD>
<META content=text/html;charset=iso-8859-1 http-equiv=Content-Type>
<META content='"MSHTML 4.72.3612.1700"' name=GENERATOR>
</HEAD>
<BODY bgColor=#ffffff>
<DIV><FONT color=#000000 size=2>All,</FONT></DIV>
<DIV><FONT color=#000000 size=2> RDC closed just below
my 8 3/4 stop yesterday. If it doesn't trade above 8 3/4 early this
morning, I'll close the position. I hate to do it since I really believe
there will be a lot of money made in this issue over the next few years.
However, I'm not going to fight my system and I can always get back in again
later.</FONT></DIV>
<DIV><FONT color=#000000 size=2></FONT> </DIV>
<DIV><FONT color=#000000 size=2>JimG </FONT></DIV></BODY></HTML>
</x-html>From ???@??? Sun Jan 09 16:15:09 2000
Return-Path: <majordom@xxxxxxxxxxxxxxxxxx>
Received: from listserv.equis.com (listserv.equis.com [204.246.137.2])
by purebytes.com (8.8.7/8.8.7) with ESMTP id OAA09691
for <neal@xxxxxxxxxxxxx>; Sun, 9 Jan 2000 14:17:10 -0800
Received: (from majordom@xxxxxxxxx)
by listserv.equis.com (8.8.7/8.8.7) id NAA25055
for metastock-outgoing; Sun, 9 Jan 2000 13:29:59 -0700
X-Authentication-Warning: listserv.equis.com: majordom set sender to owner-metastock@xxxxxxxxxxxxx using -f
Received: from freeze.metastock.com (freeze.metastock.com [204.246.137.5])
by listserv.equis.com (8.8.7/8.8.7) with ESMTP id NAA25052
for <metastock@xxxxxxxxxxxxxxxxxx>; Sun, 9 Jan 2000 13:29:56 -0700
Received: from mout00.kundenserver.de (mout00.kundenserver.de [195.20.224.69])
by freeze.metastock.com (8.8.5/8.8.5) with ESMTP id NAA27855
for <metastock@xxxxxxxxxxxxx>; Sun, 9 Jan 2000 13:40:14 -0700 (MST)
Received: from [195.20.224.237] (helo=mx03.kundenserver.de)
by mout00.kundenserver.de with esmtp (Exim 2.12 #2)
id 127Otg-0004Ck-00
for metastock@xxxxxxxxxxxxx; Sun, 9 Jan 2000 21:25:08 +0100
Received: from [212.184.144.235] (helo=pD4B890EB.dip0.t-ipconnect.de)
by mx03.kundenserver.de with smtp (Exim 2.12 #2)
id 127Ota-0006wW-00
for metastock@xxxxxxxxxxxxx; Sun, 9 Jan 2000 21:25:02 +0100
From: lists@xxxxxxxxxxx (rudolf stricker)
To: metastock@xxxxxxxxxxxxx
Subject: Re: how to improve generalization in system optimization
Date: Fri, 29 Jan 1999 20:23:31 GMT
Organization: tbus tech-consult & simulation, 80538 munich, germany
Message-ID: <36b1e43a.5278797@xxxxxxxxxxxxxxxxxx>
References: <52.521ad6fb.25a632e3@xxxxxxx> <38790080.61685384@xxxxxxxxxxxxxxxxxx> <001001bf5a05$75f21940$0bd879c3@xxxxx> <36a8e699.1525093@xxxxxxxxxxxxxxxxxx> <004801bf5aa6$395c8960$16db79c3@xxxxx>
In-Reply-To: <004801bf5aa6$395c8960$16db79c3@xxxxx>
X-Mailer: Forte Agent 1.0/32.390
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from quoted-printable to 8bit by listserv.equis.com id NAA25053
Sender: owner-metastock@xxxxxxxxxxxxx
Precedence: bulk
Reply-To: metastock@xxxxxxxxxxxxx
Status:
Hi Ton Maas,
thank you for your comments on my Excel / Windows / date problem.
On Sun, 9 Jan 2000 11:58:31 +0100, you wrote:
>Please provide more information. Please be more precise and precise in detail.
>Describe better all your actions, program details and file details.
Its not easy to do this for my "trading system", but I'll try it:
There are about 7 *.xls files ("Workbooks"?)
The 1st (io.xls) provides the input data (dax etc from 1964 up to now)
and collects the results.
Besides of this, io.xls contains a bunch of VB routines
("subroutines") that can do several things, see below.
The 2nd to 5th (cl/pl/cs/ps.xls) are very similar and contain "trading
systems" and "system optimization capabilities" for options
(calls/long, puts/long, calls/short, puts/short).
Results from these trading systems are collected in io.xls
Any of the 4 trading systems has the following functions:
a) a parametric generator of in/out signals, primarily based on some
kind of ROC (with continuous number of days) based on 5 parameters.
b) a rating system using the in/out signals from a) based on
parameters like profit, number of winning/loosing trades, amount of
wins/losses, shape of the loss distribution, etc.
c) a "generator" and a "memory" for several sets of the 5 parameters
from a), both working like a kind of "genetic algorithm" in
combination with the VB routines from io.xls to find optimal parameter
sets.
The 6th (graf.xls) provides several graphic capabilities to show the
results from io.xls and also several details from cl...ps.xls.
The 7th (cvs.xls) finally collects several results from all other
*.xls to prepare *.csv files to be used as input in MetaStock for some
kinds of post-processing.
Some additional remarks on the functions of the "trading system":
a)"Recaclulate":
For a given date and a given (optimized) parameter set the *.xls files
can be recalculated to show the results (in/ot signals, amount of
profit, trades in the past, loss distribution, etc.)
b) "Optimize":
Driven by a VB routine, the "best" parameter sets can be searched by
genetic optimization for a given date, using several "Recalculate"
steps.
c) "Time-step":
Also driven by a VB routine, the "Optimize" function can be calculated
for a sequence of weeks (defined by the appropriate dates). This can
be used for out-of-sample testing, where e.g. "Optimize" is run for
every Friday and the resulting system (parameter set) is used to
"predict" in/out signals for the next week.
BTW: To this "generalization" problem my original questions were
directed:
>What is a "good generalization result" for a trading system? Is it for
>instance a good result to get an out-of-sample profit which is about
>50% of the profit from system optimization?
>
>Are there other important techniques (beside of periodical
>re-optimization and weighting along time) to improve generalization
>capabilities of a system? - Where should I look at? - What should I
>try?
>
>Is there also a risk for overdoing the adaptive refinement and / or
>the weighting along time? - What could be seen as a relevant criterion
>to find the right measure, e.g. for the "update frequency" and the
>ratio of maximum / minimum weight?
>
And again: Any helpful hint is welcome!
To give some inputs to your detail questions:
>1. "my system tests in Excel"
>Ok so you run Excel, and in Excel you have Excel run your "system test".
>1a. What do you mean by a "system test" ?
Nearly the same thing like a system test in MetaStock: Select a set of
parameters, run the signal generator & the rating model, analyze the
results.
>1b. Is this a worksheet, a workbook, a macro?
Don't know, if my explanations above should be more precise ... : My
trading system consists of seven workbooks, one includes a VB module
with several subroutines.
>1c. Or are your actions envoked in Windows and only calling on the Excel program
> when it is needed in your test?
No action in Windows, just starting the basic VB routine that handles
the different Excel workbooks.
>1d. Who originated/wrote your test?
Should I provide his/her name here? It's rudolf stricker.
>1e. What are his/hers programming skills?
About 30 years experience in programming in several languages, from
Algol to VB.
>2. "Changes to the _date_ are made in accordance to the date of the system test"
>What do you mean by "to the_date_" ?
The date of the Windows clock is set to the date of the "optimize"
step (see above) when the "time-step" VB subroutine walks through time
to run an out-of-sample test e.g. for 1999.
>2a. Is that to a date function in a spreadsheet's cell ?
?? I use several cells with dates in all those *.xls files to assign
the data (dax, etc.) to their appropriate date. Dates are also used
e.g. in the "time-step" routine to define a time interval and to
control the stepping process. Also some Excel-based date functions are
used for this
>2b. Is your script at that same time still in Excel/VBA?
>2c Is your script at that same time back in Windows/WinScript/VB/VBScript?
>2d. Is that envoked within an Excel script(macro), where this than will be running in
> Excel's Visual Basic for Aplications(VBA) mode (VBA in Excel is Excel dependant) ?
>2e. Is that envoked in Visual Basic (VB) (VB is a stand alone program) ?
>2f . Is that envoked in a Windows script (WinScript) (ScriptIt is a stand alone program) ?
>2g. Is that envoked in a Visual Basic Script (VBScript) (VBScript is stand alone) ?
There is no other "control language" than the VB routines (= module
with subroutines in io.xls) which handle the seven *.xls workbooks.
The VB routines are setup by some amount of macro recording in
combination with some explicit VB programming. There is no additional
stand alone VB programming, so seemingly all VBs in my posting should
be changed to VBA to follow your wording.
>3. "my Excel/VB procedures"
>Now what exactly are you running and what is running simulantainously?
>3a. VB calling on the Excel-program
>3b. VBA working in Excel
>3c. VBA working in Excel and calling on Windows?
3b is correct, I think: VBA subroutines run some Excel sheets from
different workbooks.
>4. "step through an out-of-sample"
>4a. what do you mean by "out-of-sample"?
>4b. what are these steps? in what program executed?
>4c. what is active and what is not active, eg programs, windows, scripts?
Should be more clearly now: A VBA routine ("Time-step" runs (the
"optimize" subroutine) through some loops, where the counter
represents a date to cover a time frame, e.g. 1999.
This testing is "out-of-sample", because the trading system for a
given date is optimized using _past_ data and used for trading e.g.
for the next week. This way, the "generalization" capability of the
trading system is tested, see my appropriate questions.
>5. "back-test"
>5a. what are you back testing? Data cells in Excel or a stand alone system
> that makes use of data in Excel cells?
>5b. what program(s) do you use for this?
>5c. is that also using the DDE and/or OLE functionalities of
> Windows and/or Excel and/or other program(s) ?
There is always the same answers: Some VBA subroutines in io.xls run
several Excel sheets in different workbooks. That's all. No other
programs, no other DDE/OLE functionality (explicitly programmed).
>6. "The time-of-day is untouched"
>If the time is untouched, than what commands are send to the clock
>and by whom and from where/which script?(see above questions)
I don't know, which commands are sent to the Windows (or other) clock.
Exactly this was & is my question, because I'm wondering, where the
date is changed and who did it.
The boring effect is, that the _date_ of the Windows clock is changed
to the same date, which is run by the "time-step" VBA routine, when
driving the "optimize" routine through a given interval, e.g. 1999.
The _time of day_ (e.g. 20.58h) is untouched.
Finally:
It is not my expectation to make yo analyze my complete "trading
system" down to any detail, because this would be clearly impossible,
even if I would post the complete system (about 50 megs of data) here.
I'm just interested to get some helpful idea, where I should look at
to find the source of this faulty behaviors of the Windows clock in
context to my Excel task.
But still my top-rating questions are related to the "generalization
problem":
>What is a "good generalization result" for a trading system? Is it for
>instance a good result to get an out-of-sample profit which is about
>50% of the profit from system optimization?
>
>Are there other important techniques (beside of periodical
>re-optimization and weighting along time) to improve generalization
>capabilities of a system? - Where should I look at? - What should I
>try?
>
>Is there also a risk for overdoing the adaptive refinement and / or
>the weighting along time? - What could be seen as a relevant criterion
>to find the right measure, e.g. for the "update frequency" and the
>ratio of maximum / minimum weight?
Any helpful hints?
mfg rudolf stricker
| Disclaimer: The views of this user are strictly his own.
|