[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: QQQ System Test Optimizations



PureBytes Links

Trading Reference Links

On Fri, 8 Jun 2001 13:03:28 -0500, you wrote:

>Optimization is a variant of curve fitting. 

This looks confusing to me... 

Optimization is a process to select systematically a "best" set of
values for a given parametric model to meet some given goals. So its
not a bad thing at all, and I cannot understand this "optimization
bashing" going on here.
Moreover, _any_ TA work uses optimization, e.g. selecting an indicator
from several available _is_ optimization. 

Most things criticized here in context with optimization do not deal
with optimization itself but rather with a "bad" parametric model
(e.g. robustness) or with an insufficient goal function (e.g. max
drawdown, number of loosing trades, etc).

>Curve fitting has been used for
>a very long time to analyze data. One of the caveats of curve fitting is
>that you don't use the fitted curve for conditions that are beyond the range
>of data.

Curve fitting is a fully accepted application area of general system
modeling (i.e. a combination of a parametric model, a goal function,
and an optimization procedure). And of coarse extrapolation can be a
dangerous thing, if the _parametric model_  does not cover e.g.
changes of the system behavior over time. 

So, we should not blame system optimization  for everything done badly
during parametric model and/or goal function setup.

mfg rudolf stricker
| Disclaimer: The views of this user are strictly his own.