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

Re: Indicators, Beagle or human



PureBytes Links

Trading Reference Links

A quick note on the below comments here:

Paul Tudor Jones, paired with DeMark and several large
institutions tried to do this years ago. 

They failed miserably.

Walt Downs
CIS Trading

Gary Fritz wrote:
> 
> Tim wrote:
> > If you read Gabe's statement above, I beleive he is saying that
> > given today's technology, someone with coding expertise could
> > investigate the tools I use to trade and replicate my trading
> > success and style. I am not willing to go down that road with what
> > I know about trading. By the same token, if you showed me, a
> > discretionary trader, exactly how a complex trading program or
> > indicator worked, I'm not willing to say that I could replicate the
> > system's or indicator's success. I don't know that either is easily
> > transfered from human discretionary trader to code on the one hand
> > or from code to human discretionary trader. I'm not willing to say
> > it's simply a matter of being able to 'identify, quantify and
> > systemize.'
> 
> My 2 cents:  I believe that most techniques used by discretionary
> traders COULD be "identified, quantified, and systematized."  But
> just because it's *possible* doesn't mean it's *easy*.
> 
>   * Identifying is probably the hardest part.  From my brief
>     stint in AI and expert systems, I know that the hardest thing
>     about writing an expert system is cracking open the expert's head
>     and finding out HOW he solves problems.  Very often the expert
>     has no idea how he solves the problem -- it's not a conscious
>     decision tree for him.  He just encounters a problem, and the
>     right answer pops out of the ether.  That's hard to do, and
>     it's virtually impossible to know whether you've collected
>     the "full" set of rules that the expert draws on.
> 
>   * Quantifying, if I understand what you mean, is also a major
>     undertaking.  I don't believe most discretionary traders normally
>     think quantitatively.  They don't compute moving averages and
>     RSI's in their heads.  (They may make use of indicators like
>     that, but that's a minor part of their process.)  They just look
>     at the charts and see patterns that "look" right.  That sort
>     of pattern recognition is very difficult to computerize.
> 
>   * Systematizing, i.e. programming the system you quantified in
>     step 2, is a "simple matter of programming."  Hah hah.  :-)
>     I believe you couldn't program 99% of such systems in
>     TradeStation (at least not 4.0), because of its many limitations.
>     But it can be done with suitable tools.
> 
> So, I believe it would be possible to reproduce on a computer,
> reasonably accurately, the trading process of a successful
> discretionary trader.  However, it's likely to be a HUGE project, one
> that could only be undertaken by a large trading house.  And in fact
> I suspect quite a few of them have been doing so for some time.
> 
> Gary