|
PureBytes Links
Trading Reference Links
|
At 7:43 PM -0500 12/20/01, M. Simms wrote:
>Bob - I agree, but wow, the amount of logic and complexity with this
>approach is significant....
>
>i.e., what measure to use to determine a change is required ?; how
>does that measure correlate to changes in the parameters of the
>system ?; how often should the system make a change ? How will it
>know exactly which parameters to change ? Does the system need to
>maintain multiple sets of "books" in order to be able to effectively
>revert back to it's original parameters ? Or should the initial
>parameters be foresaken forever, never to be revisited ?
A perfect adaptation can be very complex but you can often get 80% of
the benefits with very simple measures (per Pareto's Law).
Most fixed point-values of stops, etc., can be adapted by relating
them to the price or the volatility with measures such as
AverageTrueRange. The AverageTrueRange of daily bars of the Nasdaq
composite index has varied from about 20 to about 300 over the past 3
years. To expect any system with fixed limits to work well over such
a range of volatility is unrealistic.
But systems that scale the limits with volatility might work fine
over such a range. If the weekly adjustments the person was making
were really just manually adjusting for the underlying change in
volatility, the re-optimization makes sense - but more adaptive
methods would probably work better.
>All good questions.....and this approach certainly flys against the
>"simple-is-better" philosophy espoused many times on this list.
A "simple-is-better" basic idea for a trading system is great. I like
the analogy of an autopilot flying an airplane. The "simple idea" is
to stay at a fixed altitude and fly directly to the desired airport.
That works about 80% of the time. If you are happy almost crashing
the other 20% of the time that is all you need... Such a trading
system can have 5 to 20 lines of code. We see them on these lists and
in magazine articles all the time.
But if that is not good enough, you can do one of two things:
> Have a pilot in the cockpit who can judge when the autopilot
is having trouble and take manual control.
> Start adding more software to handle more and more of the
other 20% of the cases.
A "fully developed" autopilot includes code to handle hundreds of
special cases that occur only rarely. A fully developed trading
system that can adequately handle 99% of the cases with adaptation to
changing market characteristics, intelligent bet-sizing, risk
management, crisis stops, etc., might be several pages of code.
Traders never go broke when the market is in the 80% "normal"
condition. Trading then is pretty easy. It is the other 20% that gets
them. "The devil is in the details..."
Bob Fulks
|