|
PureBytes Links
Trading Reference Links
|
Hmmm, my experience leads me to believe your ideas represent the
"conventional" wisdom on optimization rather than mine. Nevertheless, you
can optimize as frequently as tick-by-tick if you want. As long as parameter
values stay in the same peak "neighborhood," it shouldn't materially affect
performance and thus re-optimizing is usually just a waste of time. But when
a parameter value goes from 100 to 10 for example, you should reconsider as
the radical change is telling you something. Either the system has cycled
into a down period or the character of the market has changed. Your response
should be to leave the system alone and rotate into another market until it
cycles back, or design a new system in consideration of the market's new
character. By character I mean a change in exchange rules, political
metteling, options trading initiated, etc.
More importantly, if doing this works for you...trade it! However, if by
"confidences" you mean the statistical measure that assumes the Normal
Probability Distribution, I can assure you that the markets do not play that
tune.
Have a nice weekend,
Rick
-----Original Message-----
From: John Scudiero <scudiero@xxxxxxxxxxxxxxxx>
To: metastock@xxxxxxxxxxxxx <metastock@xxxxxxxxxxxxx>
Date: Friday, August 14, 1998 9:00 PM
Subject: Re: Optimizing
>Rick, you answer is the std answer on how to develop a trading plan.
>
>But lets think outside the box for a monent and let me ask you the
>following.... If you could optimize the last 250 days (i.e. one years
worth
>of stk data) what would be the confidences that the optimize indicators
>would work for the next 20 days (ie one month). And then at the end of
>those 20 days reoptimize the last 250 days and repeat the process.
>
>I assure you the confidence is over 90% accurate.
>
>regards
>johnnie
|