|
PureBytes Links
Trading Reference Links
|
Simon, for discussion of your point on lag in moving averages, see TASC for
March 2000 (p. 16ff) and/or Ehlers' new book (Rocket science for traders)
Regards, Peter Riddick
----- Original Message -----
From: "Simon Roberts" <Roberts@xxxxxxxxxxxxxxxxxxx>
To: <metastock@xxxxxxxxxxxxx>
Sent: Tuesday, August 21, 2001 8:44 PM
Subject: Re: Weekly/Monthly data
> Sean,
>
> Interesting point. This reminds me of the arguments for and against
plotting
> a moving average where we commonly do so now. If considered logically,
and
> from the mathematical viewpoint of what the average represents, then the
> average should really be plotted at the point on the chart which
represents
> the mid point of the average, and not the last point, as common practice
> today, and since the beginning of charting.
>
> So for a 21 day simple MA you would plot it at the middle pont of the
> average, that is today minus 11 days.
> It of course raises more interresting questions as to where to plot
weighted
> averages, or exponentials. Presumably at the point representing the centre
> of gravity of the data, not the mid point. Not very mainstream, and not
> much help.
>
> So for weekly data, a more representative figure might be the close on
> Wednesday, or even an average of the weekly closes. However, usually a
> weekly figure is used where less information is needed, not more, hence
the
> close of the week.
>
> Simon
>
> -----Original Message-----
> From: Sean Taylor <cloudnine@xxxxxxxxxxxx>
> To: Metastock List <metastock@xxxxxxxxxxxxx>
> Date: 21 August 2001 01:35
> Subject: Weekly/Monthly data
>
>
> >Is anybody out there using weekly (or even monthly data)?
> >
> >It occurred to me that if using weekly data for a system, is it truely a
> >reflection of the weeks activity to go off the weeks close? Might it be
> >better to perform some sort of average on the weeks data to better
reflect
> >the weeks activity? i.e. the average close for the week.
> >Metastock takes the Friday close and uses this for weekly data.
> >
> >I raise this because I've come up with a system that works well on the
> >Russell 2000 weekly but not daily.
> >
> >It just seems that the data I'm basing the test off could vary as easily
as
> >an individual day.
> >
> >Thoughts?
> >
> >Thanks,
> >Sean
> >
> >
>
>
>
>
|