|
PureBytes Links
Trading Reference Links
|
Thanks for the feedback. We often use Kalman filters to combine the
outputs from more than one model into a single, final signal. Also
use it in variable parameter work. Interesting that you can use VB
and JS in AmiBroker. Wow!
Regards
MG Ferreira
TsaTsa EOD Programmer and trading model builder
http://www.ferra4models.com
http://fun.ferra4models.com
--- In equismetastock@xxxxxxxxxxxxxxx, "Ed Hoopes" <reefbreak_sd@xxxx>
wrote:
> I read John Ehlers book "Rocket Science for Traders" - he devoted Ch.
> 6 to the Hilbert Transform. Figure 6.7 has it coded in EasyLanguage.
> Looking at the code, I think it would perform like the Triangular
> option of the moving averaging formula in MS 9.0 - further, the code
> looks simple enough to code in MS-FL - which, of course, I am too lazy
> to do.
>
> My experience with Kalman is that the sources of noise in the signal
> need to be well characterized and constant. This is true for most
> electronic / electroptic applications. However in the stockmarket, I
> found thats not true. To run a Kalman filter, the several terms
> required to calculate the next days change need to be constantly (
> every few months) re-tuned. After tuning, things work OK for a while,
> then need to be tweaked again.
>
> By the time you have a term for every thing you think impacts the
> price, there are many functions and look up tables. You need look up
> tables, since noise sources in the stock market are seldom Gaussian
> like Kalman wants. They also suffer from the "Fat Tail" issue -
> namely that extreme outliers are much more common that would be
> predicted by a Gaussian distribution.
>
> For this reason, I no longer use my Kalman filter, despite many hours
> and hundred lines of JavaScript code it required (you can run JS and
> Vis Basic - both free - inside of Standard AmiBroker.)
>
> --- In equismetastock@xxxxxxxxxxxxxxx, mgf_za_1999 <no_reply@xxxx>
wrote:
> > For an easy introduction to the Kalman filter, download the PDF
> > article from this page.
> >
> > http://www.ferra4models.com/Kalman.html
> >
> > Ed, given your background, have you ever used the Hilbert transform?
> > It is supposed to give an idea of both the length and the strength of
> > the trend, if used correctly.
> >
> > Regards
> > MG Ferreira
> > TsaTsa EOD Programmer and trading model builder
> > http://www.ferra4models.com
> > http://fun.ferra4models.com
> >
> >
> > --- In equismetastock@xxxxxxxxxxxxxxx, "Ed Hoopes" <reefbreak_sd@xxxx>
> > wrote:
> > > The problem that you are trying to solve is to extract some
> > > information (trend) from a noisy signal. My experience ( which
comes
> > > from chemical process control and stock trading ) is that the
simplest
> > > SMA or EMA will extract about 85% of the information content from a
> > > noisy signal. So the improvements you seek by going to more exotic
> > > calculations will not improve things very much.
> > >
> > > My goals for trenders, are:
> > > Fast response to trend changes
> > > Low rate of whipsaws
> > > As you can see, these tend to be mutually exclusive.
> > >
> > > The categories are:
> > >
> > > Moving Averages:
> > > SMA's have a constant weight for each term of the average
> > > Variable weights for each term - EMA, WMA, Gaussian, etc
> > > Weights that change depending on some market condition - like
> > > volatility( try ATR ) - Adaptive moving averages
> > >
> > > Filters:
> > > There are many of these (mostly coming out of the electronics
> > > industry) like a Kalman filter. Kalman takes the previous value of
> > > the smoothed function and adds some fraction of the current value of
> > > the signal based on a complex formula. (Kalman filters are used to
> > > position disk drive heads and guide smart bombs) You can look in
> > > Google for Kalman filters, and will discover they are difficult
> > > mathematically. Also you will discover that MS - FL is entirely
> > > inadequate to implement these - Better to use AmiBroker software -
> > > cheaper and MUCH more powerful programming language.
> > >
> > > Curve fitting:
> > > Here you fit some mathematical function to the price data. MS
> > > has a linear least squares fit, but higher order fits work much
> > > better. (don't try programming a quadratic or cubic fit in MS - FL)
> > > Fourier transforms are also used, but mathematically complex.
> > >
> > > ______________
> > >
> > > My experience has been that there is some gold to be mined by trying
> > > to improve your trend following indicators, but not very much. My
> > > production trading system uses custom coded adaptive moving average,
> > > and curve fitting.
> > >
> > > Cheers,
> > >
> > > Ed Hoopes
> > >
> > >
> > >
> > >
> > > --- In equismetastock@xxxxxxxxxxxxxxx, "tan ming"
> > > <reminiscenostalgia@xxxx> wrote:
> > > > This post hope to starts a discussion on what are the available
> > > methods or
> > > > indicators which can effectively used to determine whether current
> > > price is
> > > > in a trend mode. Personally moving average is my main indicator to
> > > dtermine
> > > > trend. However as many of you guys know, it suffer when the
> market is
> > > > starting going to a tradning range. Pls join in the discussion
> on the
> > > > following topic:
> > > >
> > > > 1) What are the effective methods or indicators to determine
whether
> > > the
> > > > current price is in a trend mode.
> > > >
> > > > 2) What are the effective methods or indicators inorder to tell
> > > whether the
> > > > current price in the early, middle or mature trend?
> > > >
> > > > Great guys like Jose, Roy, MG and many of them in the forum pls
> > > offer some
> > > > pointers.
> > > >
> > > > _________________________________________________________________
> > > > Block pop-up ads with MSN Toolbar. http://toolbar.msn.com.my/
------------------------ Yahoo! Groups Sponsor --------------------~-->
<font face=arial size=-1><a href="http://us.ard.yahoo.com/SIG=12hoksc6b/M=362343.6886682.7839641.1493532/D=groups/S=1705375617:TM/Y=YAHOO/EXP=1123575510/A=2894350/R=0/SIG=10tj5mr8v/*http://www.globalgiving.com";>Make a difference. Find and fund world-changing projects at GlobalGiving</a>.</font>
--------------------------------------------------------------------~->
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/equismetastock/
<*> To unsubscribe from this group, send an email to:
equismetastock-unsubscribe@xxxxxxxxxxxxxxx
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|