[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Fwd: improvements for "catscan" & other systems



PureBytes Links

Trading Reference Links

This was sent to me privately and I asked for Bob's approval to post it to 
the list.  It is a good plan for improving systems.  Hope it helps or 
generates ideas in your trading.
Lynn
Return-Path: <bfulks@xxxxxxxxxxxx>
Received: from  rly-yc03.mx.aol.com (rly-yc03.mail.aol.com [172.18.149.35])
	by air-yc04.mail.aol.com (v59.4) with SMTP; Sat, 08 May 1999 15:13:16
	-0400
Received: from pm05sm.pmm.cw.net (pm05sm.pmm.cw.net [208.159.98.154])
	  by rly-yc03.mx.aol.com (8.8.8/8.8.5/AOL-4.0.0)
	  with ESMTP id PAA11172 for <MRLYNNG@xxxxxxx>;
	  Sat, 8 May 1999 15:13:16 -0400 (EDT)
Received: from [166.62.75.205]
 (usr2-dialup25.mix2.Boston.cw.net [166.62.66.89]) by PM05SM.PMM.CW.NET
 (PMDF V5.2-29 #35323) with ESMTP id <0FBF00DUJHDXO9@xxxxxxxxxxxxxxxxx> for
 MRLYNNG@xxxxxxx; Sat,  8 May 1999 19:13:15 +0000 (GMT)
Date: Sat, 08 May 1999 15:12:54 -0400
From: Bob Fulks <bfulks@xxxxxxxxxxxx>
Subject: Re: improvements for "catscan"
In-reply-to: <15522502.2465dca4@xxxxxxx>
X-Sender: bfulks@xxxxxxxxxxxx
To: MRLYNNG@xxxxxxx
Message-id: <v04020a06b35a3b219d03@[166.62.75.205]>
MIME-version: 1.0
Content-type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit

>Your post was one of the best that I have seen on this topic of analyzing and
>improving a system.  A few weeks ago I tried to start a thread on Refining a
>System but it go little response.  You appear to have a good insight into
>system development and alternative ways to code the various methods used for
>trading.

Hi Lynn:

I recall seeing your request. It was on the Code List, not this one. I
thought about replying but was too busy at the time to do it justice.

I have a fairly formal process I use to improve a system.

I first try to determine what the basic idea is. What are they using from
the past that they think will correlate with price in the future.

I then try to isolate that characteristic and see if it in consistent. This
uses the advantage of hindsight so it is pretty easy to do.

If this seems to show some consistency, I then try to figure out how to
accomplish this in real time. This is harder.

Attached below is part of a recent post on designing a trading system. It
give you some of the ideas. You might also see my analysis of the T0413
system on the Code List on 4/18/99 "Subject: Re: CL_A GOOD SYSTEM??? -
T0413"

Hope this helps.

Bob

-----------------------

But, back to indicators, I now use them in a different way. To build a
trading system, you need to find something that happened in the recent past
that correlates fairly well with what the price is going to do in the
future. This is fundamental. Without it, this is all just gambling. So I
use indicators to search for characteristics that might be tradable.

Examples might include:

  - Tops seem to occur about every 15 bars

  - Prices tend to move up after a "V" shaped bottom

  - Prices seem to increase when they cross over the 50-day moving average

  - Prices tend to bounce off of "support and resistance" lines

  - The floor/specialist tends to run stops before a change in direction.

You need to search for these in historical data and indicators are very
useful in locating them.

Then, once you spot some characteristic that you think might be tradable,
you need to test how consistent it is in the historical data; how reliably
does it signal a price move. This is usually pretty easy after the fact.

If it seems to be pretty consistent, you then need to find a way to locate
the characteristic as it is happening. I have found that the traditional
indicators are basically useless for this. By the time they issue a signal,
the thing you are looking for has passed. This is because to look smooth
enough to interpret by eye, indicators need to use a lot of filtering which
introduces lag.

I almost always end up working directly with the price to find the
characteristic I am looking for in real time. I often use digital filters
to spot the event well before it becomes apparent to a common indicator,
(quite possibly because on my training as an electrical engineer.) A simple
example of such a digital filter was a trading system published by Bob
Brickey about two years ago (code slightly simplified by me):

------

{ rpbTinyNN System - Last Modified 3-13-97 }

vars: Sum(0), NN(0);

Sum =
      + .1904762  * Close[5]
      + .047619   * Close[4]
      - .0952381  * Close[3]
      - .40476197 * Close[2]
      - .0476191  * Close[1]
      + .3095237  * Close;

NN =  1 / (1 + ExpValue( -Sum )) - .5;

if NN > 0 then
   Buy at Close
else if NN < 0 then
   Sell at Close;

-----

It is not very apparent to a person what this does but a computer has no
problems trading it.

Often I use some of the techniques that form the basis of the popular
indicators. For example, The Stochastic indicator is based upon the
"discovery" that the position of the closing price in a price bar tends to
signal a change in direction. It is useful to understand this so it is
worthwhile knowing why various indicators work. But rather than waiting for
the filtered lines of the Stochastic indicator to start moving, I would
design some computer technique to locate this characteristic directly.

This is not to say my way is the only way. Different people use different
techniques to develop trading systems. I happened to like such so called
"mechanical" trading systems that do all the thinking for me. I find that I
cannot objectively trade very well by looking at charts and indicators. But
there are many "discretionary traders" who are very good at it. In fact, I
suspect that the best traders are discretionary traders. Their mind has
learned to handle far more complicated cases than a computer program ever
can.