--- In [email protected], "Howard B" <[EMAIL PROTECTED]> wrote:
>
> Hi Rich --
> 
> Do I understand correctly --- the RAR remains about the same for
many runs,
> each of which had some random noise added to the input price data?
> 
> If so, that is generally a good sign -- your system is recognizing and
> giving buy and sell based on the signal portion of the data, as you
asked it
> to, rather than the noise portion.
> 
> As always, be sure to perform tests on out-of-sample data that was
not used
> to develop the model.
> 
> Thanks,
> Howard
> 
Howard,

Yes, there's 40 runs taking the noise weighting from 0 - 1. And the
RAR is hardly affected (other than scatter). But the number of trades
(exposure) drops of drastically - which I'd expect......I think?

The RAR when performed in sample is about 150%, this drops to between
75 and 100% OOS. This again remains consistent as it walks forward -
so there is drop off there but the figure still seems good - again
it's understanding the context - how much drop of is expected (or
should I ask how long is a piece of string)?

Your input is greatly appreciated - sadly I don't have a lot to offer
in return (other than another book order...:-) but surely others will
also benefit from your replies.

Thanks again,
Rich




Reply via email to