|
PureBytes Links
Trading Reference Links
|
be careful......you're trying to apply a linear control process
(negative or error correcting) closed feedback loop to a nonlinear
system. linear may be in a special case of nonlinear. i'm thinking
about when an predictable event (fed announcement) produces a chaotic
and unpredictable reaction (resonating price shocks in bonds or s&p).
depending upon the frequency response of your controller (lag), the
command law you're trying to model, and your error correcting feedback
loop to the input, you might inadverently blow up your system by
setting up over- and undercompensating cycles triggered by the spurious
events. i've seen it happen many times in materials testing using
automated servomechanisms.
depending on your time frame, you might want to prefilter your data
with a kalman filter and eliminate much of the noise. or better yet,
formulate a kalman filter that adapts over time to changes in the noise
characteristics. that's assuming, of course, that you're dealing with a
linear system.
TJ
maybe we can mosey over to the ati list and pursue this further....i've
been thinking about this since 92.
--- Val Clancy <valclancy@xxxxxxxxxxxxx> wrote:
> The reason why I picked control theory to read up on is that
> parts of it contain some important concepts about adaptive
> feedback systems, such as open loop and closed loop feedback
> systems. I strongly believe that the next step in trading systems
> design will be adaptive systems. Some of those basic control concepts
> are really key concepts for designing any adaptive system.
|