|
PureBytes Links
Trading Reference Links
|
>Subject: Re: EL precision can cause problems
> Date: Tue, 06 Jul 1999 19:03:53 -0400
> From: Robyn Greene <greene@xxxxxxxxxxxxxxx>
> To: rbl@xxxxxxxxxxx
> CC: omega-list@xxxxxxxxxx
>
>I'm always reluctant to get into these discussions because I'm not sure that my
>SuperCharts works like TS. Anyway - at least in my SC - there are other math
>problems. For example - if you run a system that has a 15 period indicator - and a
>60 period indicator - you have to go 60 bars back to make the system run. But
>frequently (not with all indicators - but with quite a few) - the 60 bars back will
>give you a different result on the 15 period indicator than you'd get if you ran
>that indicator 15 bars back. I never did figure out a fix for this problem. Robyn
>
This problem is most likely related to MaxBarsBack (MBB). One of TS's
shortcomings is the
way it accumulates variables. During the MBB period it saves the value
of variables that
are based strictly on price or volume. However,variables based on other
variables are not.
So when the indicator begins plotting (at the end of the MBB) some of
the variables are 0.
Only after the MBB period going FORWARD do all the variables become
valid. So your 15-period
indicator is probably not valid until the 30th bar, which is why you get
a different value
using it with 60 MBB.
This has implications for systems, too, since early signals may not be
valid. This is
especially true if using long MBB.
To fix the problem simply inhibit plotting or buying/selling until the
bars equal double
the MBB.
Also, recursive indicators (those that reuse their previous value, such
as exponential
moving average) require a long period to stabilize.
donc
|