|
PureBytes Links
Trading Reference Links
|
Rudolph:
I think the theoretical accuracy will increase if you use an appropriate
distribution (versus historical results). The problem, though, is will the
distribution you choose -- whether Normal, leptokurtic or whatever -- match
future trade results? Even if it matches future trade results for the next
month, the distribution will probably change again the following month.
This business is not an exact science.
What distribution to use, is the $64,000 question. Until that question is
answered, we just have to limp along with the tools we have and a firm stop
loss.
Regards.
----- Original Message -----
From: rudolf stricker <rst@xxxxxxxxxxx>
To: <metastock@xxxxxxxxxxxxx>
Sent: July 17, 1999 09:05
Subject: Optimal f and analytical probability distributions
>
> Having a closer look at the Optimal f procedure (as posted here by Ed
> Winters) I have some doubts, if this simple "integration process"
> along the historical trades will lead to sufficient accuracy in terms
> of probability, when it comes to "tailed" distributions (e.g. for
> short positions in options). At least for the system tests I did, the
> number of points in the tail was far to low to get sufficient accuracy
> for this "rectangular" integration process.
>
> So my question is:
> Will the accuracy of the probability really increase, if I use an
> analytical probability distribution as an approximation for the
> discrete trading events?
> And if so: What type of probability distribution is most appropriate?
> Should I use e.g. Weibull's distribution? - Or any suggestions?
>
> BTW: Increased accuracy of the "tail-losses" (e.g. probability xx of
> loosing more than yy%) imo could be very helpful as an additional
> criterion in system optimization.
>
> mfg rudolf stricker
> | Disclaimer: The views of this user are strictly his own.
|