|
PureBytes Links
Trading Reference Links
|
Dimitris:
Yes, technology is getting faster and faster.
But to what end? Does it make sense to go down the continuous
optimization route? In my experience, that route does not work any better
(maybe worse) than just settling on a single set of parameters, learning how the
indicator behaves, and using it. Perhaps it is just my experience,
but system robustness seems to decline as optimization increases.
Seems to be the nature of the beast.
Bill
----- Original Message -----
<BLOCKQUOTE
>
<DIV
>From:
DIMITRIS
TSOKAKIS
To: <A title=amibroker@xxxxxxxxxx
href="">amibroker@xxxxxxxxxxxxxxx
Sent: Wednesday, July 25, 2001 9:05
AM
Subject: [amibroker] Re: Oversold and
Overbought Time
Dear Bill,Thank you for your reply.There is,
perhaps, a misunderstanding here, referring to this "glutton for
punishment".In the case you do not know, the scan of the whole market
takes 22 seconds, with the use of an 800 Mhz PIII for 385 stocks for 390
days.On the other side, read please the few lines of #3341.It took ~5
seconds.I will revert with details on the subject.Dimitris
Tsokakis--- In amibroker@xxxx, "wavemechanic" <wd78@xxxx>
wrote:> Dimitris:> > I think you are going down the
overfit/overoptimization road, and forcing the OS and OB time to be equal
is as arbitrary. Why not just optimize both, thereby letting the
market tell you what it wants. Of course, if you are a glutton for
punishment then you could do it for each issue, and recheck weekly, etc.
as it will undoubtedly change over time.> >
Bill> ----- Original Message ----- >
From: Dimitris Tsokakis > To: amibroker@xxxx
> Sent: Monday, July 23, 2001 6:11 AM>
Subject: [amibroker] Oversold and Overbought Time > >
> Oversold and overbought levels are usually selected by
experience and may> be not satisfactory for a certain
Market.> We can change levels in order to give a more
reliable description of the Market.> > 1. When
we select 30, 70 , i.e. equal distance from 0, 100, we have not any
reason> to do it. Perhaps 24, 70 for example would be more
realistic.> > 2. I have the following idea:
> I will examine the oversold time for each stock, I will
take its average and so I will> define the "mean
oversold time (MOT)" for the Market.> This will be done for
an oversold level which gives MOT >5% of total days, else oversoldis
meaningless.> > 3. Then I will search for a certain
overbought level which gives the same "mean> overbought
time". I consider this more fair for the Market, whereas 30, 70 or 20,
80> sounds abritary.> In other words, I
ask levels which share the time equally for oversold and
overbought> phases.> > Any
opinion on this ?> (The thought behind the curtain is that
buyers and sellers wait nearly the same > time
interval, until they change the trend.)> (formulas are
almost prepared, I want to discuss the basic thought).>
> Best regards> Dimitris
Tsokakis> >
Yahoo! Groups Sponsor >
> >
> Your use of Yahoo! Groups is subject to the Yahoo! Terms
of Service.Your use of Yahoo! Groups is subject to
the Yahoo! Terms of
Service.
|