[FoRK] Science without explicit theory

Jeff Bone jbone at place.org
Sun Jul 6 20:05:01 PDT 2008


On Jul 6, 2008, at 9:23 PM, Aaron Burt wrote:

>> Ah, rather like a MIMO control loop, where you have multiple inputs
> (including feedbacks) that each affect multiple state variables.  Only
> more complex, of course.  Using Control Theory, one could maintain a
> plant-model of the ANN, but then you'd just have a trivial linear  
> model
> of the gross behaviour, useful only for short-term predictions of what
> inputs are needed to get a particular output.

Mmm hmm.

> (Though "Memoirs of a Stock Operator" does come to mind when I see the
> size of some of these hedge funds.  Give 'em time and some of these
> automated-trading systems are gonna start playing pricing games.)

You think this doesn't happen already?  You think that's AIR you're  
breathing now?  Hmm.

The dirty secret, though, is that most of the larger hedge funds are  
simply exploiting cost structure and other aspects of market macro-  
and micro-structure that so far require a lot more capital, privilege  
and placement to make workable.  It's all stuff that goes naturally  
goes away, a race to the bottom, assuming market operators themselves  
are rational and uncorrupted;  it's unsustainable advantage.

There aren't many out there that are really making the huge dough  
based on actual prediction technology.  The market's a random walk,  
isn't it?  Hell, even The Prediction Company went on an excursion into  
different business models (e.g. fraud detection) before being bought  
by UBS.



jb



More information about the FoRK mailing list