|
PureBytes Links
Trading Reference Links
|
Michael,
Optimization is not the key to making money; it is the information about
parameter stability, which is the only way to use it. Too many
parameters do not make it better.
When testing remember to use money management and once you do that you
will see how results will change once using REAL portfolio level money
management. When using software that considers that you will see what I
mean.
Finally you ask "for the right direction for this analysis; finding the
software and the right person". Try WLD3 because it can do the above
mentioned. Good luck.
Kind Regards,
Volker Knapp
(www.wealth-lab.com)
||-----Ursprüngliche Nachricht-----
||Von: Michael Tepper [mailto:michael.tepper@xxxxxxxxxxxx]
||Gesendet: Thursday, April 22, 2004 6:14 PM
||An: Omega-list@xxxxxxxxxx
||Cc: MEE
||Betreff: analysis of trading systems
||
||I would appreciate any suggestions from the members of the group,
||regarding the below, for the right software, analytical approach, and
||good consultant.
||
||We want to do analysis of our trading systems. We are looking to
||analyze 10-50 variables used in each of 10-30 stock trading systems.
||The goal would be variable set selection.
||
||We are looking for the right direction for this analysis; finding the
||software and the right person who can help us with the analysis.
||
||I can describe what I think the goal is (but maybe it can be better
||stated): I think the goal is how to analyze our simulations such that
||we choose a good variable set (varset), and increase the odds that we
||will realize real profits, not just simulated profits. How do we use
||our simulations to choose good systems to trade?
||
||Example: We have tests from an optimization on a list of stocks.
There
||are multiple variables (say, two dozen variables in a test). The
||optimization tests show varying results.
||
||Some possible questions: How is each variable affecting results (i.e.
||the relative impact versus other variables)? How to ascertain the
||confidence level for a variable selection (what we call a varset)?
Can
||we develop a confidence level for each varset and compare the
confidence
||levels to each other? Is that useful for the real world? Is there an
||objective way to do all this? What questions we should be asking of
the
||data? What is the way to approach the test results?
||
||Thanks.
||
||Michael Tepper
||
|