|
PureBytes Links
Trading Reference Links
|
On Fri, 15 Dec 2000 11:44:46 EST, you wrote:
>Optimization should converge on the flat part of the multidimensional
>parameter space.
There is much talking about this _flat_ part (whatever that may be),
but this sounds rather strange to me:
If there is a rather flat optimal area in an optimization function,
then the optimal area is rather large.
If there is a large optimal area (i.e. large market imperfection),
then it would be identified easily by many market actors, and it would
no longer exist.
Therefore, having _flat_ optimal areas in trading systems imo is
something like wishful thinking (beside of the general difficulty to
define "flat" in this context).
>Presently, I'm looking at mathematical methods to automate this. Currently
>it is a combination of art and science.
The "best" optimization procedures for multi-extremal (i.e.
"complicated") optimization functions imo are from the family of
"genetic algorithms", preferably in the context of "genetic
programming". Simple versions can be easily setup e.g. in Excel.
mfg rudolf stricker
| Disclaimer: The views of this user are strictly his own.
|