|
PureBytes Links
Trading Reference Links
|
Dans un courrier daté du 23/09/98 19:31:06 , Phil Lane écrit :
<<
I've spent many years studying the market with TS. I've tried literally
thousands of great ideas that DON'T WORK. Maybe I'm asking too much. The
goal is to find something robust, consistent, etc. that works across ALL
CONTRACTS going back to the beginning. (As far as I'm aware this was 1982
for the SPs??) There must be a large sample of trades, with a good
distribution of profits (doesn't depend on one grand slam trade). Signals
without an inherent up or down bias. An equity curve steadily ramping up
year after year.
>>
I have done the same than you during years.
With the same deceptive results and a big dustbing full of worthless systems
(that I ever considered to be great before coding and testing...)
There are some things that fairly works on mostly all markets (I nevr said
all, but i do not work anymore on dailydata), and they are not ready coded in
any software, TradeStation included.
The best thing you can do is to try some recent AI software like Neuroshell
Trader or Biocomp Profit.
If I'm correct, these new NN package should be able to better exploit the few
indicators that you may have found worth to keep on your HD.
The solution is, from my experience, somewhere around this if you do not want
to waste your time.
But beware of overfitting.
As you use to say, I have no affiliation with Neuroshell Trader (Ward Systems
Group) and Biocomp Profit.
Just wanted to pass the information, but I have not tried these products.
No affiliation, of course, and you know why...
<<
Evidently there is a large random element, since practically nothing in the
Deterministic system arena will hold up under the light of backtesting.
I've long wondered if there is a way to treat things essentially as random,
flipping a coin for long or short. The essence of the system would then be
to manage the trade according to some clever profit/loss plan. Has anyone
tried such an approach? Is it viable? Is it patented?
>>
I remember of something like this published in an old Club 3000 issue.First
testing showed it viable, further not in a later issue due to huge and
unacceptable drawdowns observed.
I do not think that it was patented...
<<
Apparently there also is a very small deterministic element. Over the years
I've found just a couple of things that really do have a good statistical
probability. Toward this end the market will show you what to do, but only
if you ask the right questions. And the solution may not be something you
feel "comfortable" with, not the way you think trading should be. But the
market doesn't care about that. If you want some cool charts to look at,
here's my latest web spew: http://www.accessone.com/~logical/CPmain.htm
(sorry, haven't quite got the System Stats page together yet).
>>
There are some deterministic elements, but are to be considered in
association (this is difficult to conceptualize in classical programming
because out of the scope of our brain structure that is more intuitive than
truely analytic).
(snip of the trailing stop discussion and usual topics CNBC and color glossy
ads problems because all of that has already ben said).
FYI, when I sell a TradeStation copy, I often insist on the fact that its only
a tool and that what you buy here is an empty shell that allows to write
anything you want with some known limitations and known workarounds.
Curiously, I does not frighten the customer to buy the product...
Sincerely,
-Pierre Orphelin
|