|
PureBytes Links
Trading Reference Links
|
> The time required to complete this process seems rather excessive,
> five hundred hours. Is there any way to speed up the back-testing
> process,
Yes. Optimize less.
If you're crunching on an optimization for TWENTY DAYS, you are doing
WAY too much optimization. Either you are optimizing one or two vars
to a sub-atomic level, or (more likely) you're optimizing a whole
batch of variables all at once.
That's a great way to overfit a system and end up with terrible out-
of-sample (and realtime) performance. Do you understand the dangers
of curve-fitting? How many variables are you trying to optimize in
this beast?
If you have a large number of parameters (like, more than two, maybe
three at the most) that are tightly interdependent, you need to
rethink your system design. Try to combine logic so you don't need
to tune so many inputs. If your parameters aren't interdependent,
then don't optimize them all at the same time.
Backtesting and optimization are powerful tools. Used correctly,
they can help you get the best performance from your system. Used
incorrectly, they can fool you into thinking you have a great money-
maker, when in fact it's going to disintegrate as soon as you try to
trade it in realtime. Make sure you understand the strengths,
weaknesses, and dangers of the tool.
Gary
|