|
PureBytes Links
Trading Reference Links
|
On Wed, 20 Dec 2000 21:33:57 -0600, you wrote:
>I wouldn't say that. I think any ideas presented well are of interest here. Heck,
>most of the traffic on this list is noise and of little value. I'm interested in
>hearing what applications you "work in".
Because I'd not like to add too much to this "noise", I think that
it's not the best place here to talk about my activities in product &
process design, where I try to concentrate on the _know-how_, which
imo represents the most valuable (but often wasted) resource in our
days.
>Does "determinism" imply predictability? or does it simply mean "non-random"?
Imo, "non-random" is (nearly) the same like "deterministic", even if
we don't have valid models for all deterministic processes to make
helpful predictions.
>I'm interested in this. Please forward a copy.
Hope, you've got it.
>my background is chemical engineering. So, although I can probably
>understand your paper, I wouldn't feel qualified to critique it from a "financial"
>perspective.
Nevertheless, any critics, comments, questions, etc are welcome.
>I believe, that the performance of most systems
>developed in the public domain is no better than random. The results people "see",
>the performance statistics, are simply small samples from a random population.
There might be (at least) three reasons for this, which are wel-known
in system identification and optimization:
First: The (trading) model should represent a sufficiently large and
accurate segment of the real system, i.e. the market.
Secondly, optimization should be done using an _appropriate goal
function_, not restricted to profit only.
Third: System parameter optimization should be done with
"generalization" in mind to support predictive capabilities.
mfg rudolf stricker
| Disclaimer: The views of this user are strictly his own.
|