Forum Archive Index - March 2004

["tennyson@caverock.net.nz" <tennyson@caverock.net.nz>]

Hi trader 100,

>
>Academic finance is based around the concept of Market Efficiency.
>"Weak Form Market Efficiency" postulates that abnormal profits can't
>be made using widely known information such as past prices.
>

I am not sure what time period your comments come from, but I would
seriously doubt that 'academic finance' today gives much credence to
the efficient market hypothesis.  IIRC the theory has even been
discredited by those who originally proposed it.

>
> Academic
>papers that have tested technical analysis have found that it is not
>possible to make profits that are statistically significant greater
>than a buy and hold profit after transaction costs are accounted for
>(with a very few exceptions).  These papers have focused on simple
>moving average rules.
>
>My personal trading suggests that TA has a lot of value but I tend to
>include a lot of visual interpretation in my trading systems which
>make them difficult to quantify and test.
>

I hope Travis Morien from the aus.invest forum won't mind me quoting
him, but here is what he posted on 4th March when discussing this
very
subject.

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

Humans are highly adept at seeing patterns that they are
looking for.  We've evolved brains that make us highly perceptive to
things like a large dangerous animal slinking towards us in the long
grass, but the problem is we get many false positives – we jump at
shadows and tend to see predators more often than there actually are
predators there.  This is, from a survival point of view, a harmless
adaptation.  We've evolved to see patterns that aren't there because
we're far more likely to be killed by a false negative than a false
positive.

Hence, if you're looking for support and resistance you'll see it
everywhere.  If you are looking for well behaved trends that can be
profitably traded by a simple trendline break system then you'll get
unlimited candidates.  Try to program these systems into your
Metastock though and the computer won't see them.  The computer is
more objective than a human is.

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

>
>For an academic TA paper to be accepted it has to include
>a rule that is easily expressed so that it can be
>rigorously tested without any subjective human
>interpretation.
>

Has any serious T/A guru ( I leave out the guys who are selling 'black
box' software as obvious fraudsters ) ever claimed they use a simple
mechanical system without any human judgement?

>From what I know about T/A practitioners, most watch their portfolios
like hawks and would feel very nervous about leaving their holdings
for days on end away from their watchful eyes.    If they didn't use
their own judgement on top of whatever 'data system'  they operate,
they wouldn't behave like that.   What is it that makes you think a
simple mechanical system could exist that has universal application?

>
> I do wonder if someone would be willing
>to provide me with a basic mechanical system that they believe is
>profitable.  I have the code developed to rigorously test such a
>system.
>
>Obviously, the code would need to prove profitable on a range of
>different stock / exchange rate series to overcome the criticism of
>curve fitting.
>

If T/A is based on the weighted average of what  humans perception of
an investment is and has been, does not your above 'restriction'
effectively mean that the way a human perceives the market behaviour
of all (shares)/(exchange rate) data must be the same if your 'silver
bullet T/A algorithm' is to work?     This seems very unlikely.

It seems your restriction is designed to make your experiment easy to
perform, rather than being a serious attempt to mimic what T/A
practitioners do in real life.    If so, it is difficult to see what
you could prove from such an experiment.

>
>The gulf between academics and practitioners in their views on TA is
>huge.  Any academic papers that attempt to bridge this gap are
sought
>after.  If someone does provide me with code that stands up to the
>profitability tests and gives me permission to use it in a paper I
>would be more than happy to include them as a co-author. I look
>forward to some replies.
>

I suggest you need to seriously rethink the design of your experiment
to reflect the real world.     Good luck though.

SNOOPY

--
Message sent by Snoopy
on Pegasus Mail version 4.02
----------------------------------
"You can tell me I'm wrong twice,
but that still only makes me wrong once."


----------------------------------------------------------------------------
To remove yourself from this list, please use the form at
http://www.sharechat.co.nz/chat/forum/