|
PureBytes Links
Trading Reference Links
|
Nice piece Craig. I agree with most of what you say.
I'm not clear on your comments about robust behavior over
"many different markets".
If you are referring to live cattle vs the emini or some stock vs
another
and you are dealing
with price(only), then possibly. I don't know, I don't trade multiple
markets. Nor do I work with only price.
If, however, you are referring to different markets as say sp data
before, during, and after "the bubble" or even the split in '97, or
any condition that changes an individual market's behavior then
I would agree.
I would imagine that anyone who has been at this for a long time
(developing mechanical systems) has, developed an
intuitive sense for what has the possible potential to be robust
even as you are writing down the line of code. To not even
consider elements that are not adaptive is a big timesaver
and constantly evolving.
To each his own but for me, forward testing puts it all together.
i.e. I've been letting a daytrading system run on a live feed since
mid-June and intend to let it run further. This ties up the loose ends
in preparation for trading.
Loose ends being....
Is the system still making decisions at market points similar to those
in historical testing that I deemed acceptable? And do they still feel
acceptable in realtime?
Is the system more stressful(or less) to trade than I had perceived?
...data outages, drawdowns(both timewise and dollarwise), points
of entry, frequency of points of entry, etc.
I'm not big on varying position size. Instead, I determine the minimum
number of contracts acceptable to make trading the system worthwhile
during slow periods. If, after any contract period, the system fails to
meet
minimum requirements, it will not be traded the next quarter.
Some might say that I am indeed varying position size by scaling back to
zero but I'm not. That exact system will never be traded again. Unless
elements are successfully added to bring it up to my standards, I'll
move on.
|