8-4-2006
Welcome and hope you find these files of use to you
This zip files includes the following:
Guppy01b.afl &
Guppy01b Index.afl & Guppy01b Diff Index.afl 7/28/06
Ken CLose-
Guppy Index and Guppy Diff Index.
You might find these interesting and useful. No, I have not done any backtesting
(yet). These work together (visually) so I suggest lining them up in three
panes in their own tab. The long term diff index (red) sort of looks
like an interesting ADX trend indicator.
I have not figured out if these are worth anything on top of the regular Guppy
MMA plot, but it seems like there is potential here.
Hull MMA.afl
7/30/06
Ken CLose-
Hull MMAs: in my searching for
more on Guppy, I kept coming across references to Hull MMas on the same pages.
Of course, a Hull MA is one which has much less lag because of the math it
employs. So I had high expectations for something "better" than the
Guppy MMAs I have posted previously. Alas, it was an unfulfilled expectation.
While the benefits of less lag might be exploited, the negatives of a much less
"intuitive" set of moving average bands take away from the visual appeal of the
MMA construction. Backtests of this and several of the Guppy
constructions do not show very good numbers in either case--but I think that
better "rules" need to be defined to wring out a possible trading system. Visual
traders can integrate all of the subjective elements for good benefit; but for
good system results, more sophisticated Buy and Sell rules need to be defined.
Fir Filters.zip
and related documents
8/1/06 Dave
Howarter-
I have been very impressed by
the results Selim has been posting of Hurst cycle analysis. I had some ideas
that digital bandpass filters used by Electrical Engineers in applications such
as audio CDs, cell telephones, etc could be used to isolate and identify the
dominant cycles of the price curve for a stock, mutual fund or index. I have set
up to methodology for doing that and am getting some good results. What I have
done is a bit complex in theory, but I have set it up so I think anyone with a
moderate knowledge of AmiBroker can make use of this.
What I am doing differs slightly from the analysis that Hurst described in his
book and analysis course. I have compared my resulting cycles with those Selim
and his colleagues are getting using Hurst analysis and my results are very
close to the Hurst analysis. I think mine require somewhat less of a visual
judgment component. Selim has graciously reviewed what I have done and given me
some comments on how he sees them comparing to the Hurst analysis he is
studying. I will attach Selim’s e-mail as selim_hurst_email.doc, as I think it
contains some interesting information about the Hurst approach.
What I have takes a bit to explain, so I have written the detailed explanation
in Hurst_cycle_FIR_design.doc and will attach that file to this post.
I also created an Excel file, cycles_prediction.xls, to project ahead of the
right hand edge of the page a projection of what the cycles are predicting. The
use of this is explained in Hurst_cycle_FIR_design.doc. A screenshot of the
output of this, based on data to 7/27/06, is attached as
cycles_prediction_7_27_06.PNG;
Unfortunately the digital filters I used would be difficult, and probably
impossible, to implement in Trade. They are very easy to implement in AmiBroker
so I have only done that and do not plan to attempt to implement them in Trade.
I will attach a file FIR_filters.zip which contains 10 different .afl files that
implement digital filters of different target periods. The use of these is
explained in the Hurst_cycle_FIR_design.doc file.
I also am attaching a screenshot of the sample output of the filters as
FIR_5_8_2006.PNG. There is more explanation of this in the
Hurst_cycle_FIR_design.doc file.
I am still in the process of developing this and testing how to use it, so what
I have now should be considered very preliminary. But the results I have to date
differ some from the Prez2 cycle predictions. The FIR cycle results changed my
viewpoint of what the market is likely to do in August and I am changing my
plans based on that so I decided to post what I have to date. I can’t guarantee
how accurately this approach is going to predict the market direction, but here
is what it is currently predicting.
- The market will be choppy and mostly move sideways until August 17, 2006.
Longs and Shorts will both find it difficult.
- Starting about August 18, 2006 the bias will shift to be much more positive
and a fairly strong uptrend will start that will last for at least a couple of
weeks.
I think there is a good potential for extrapolating from what I have done so
far, so I would be very interested in ideas from anyone else who tries what I am
posting.
FIR.vbs & FIR 800
400txt 8/1/06
Fred Tonetti-
The first step in this
process would seem to be to find a way to automate the most manually intensive
piece i.e. the generation of the coefficients and in turn the AFL ...
The SciLab site appears to be reachable today. I downloaded the software there (
v4.0 ) ... so from this point on if you are running an earlier version there may
be some differences ...
In playing with the program and looking at the doc etc. the program that runs is
WSciLex.exe i.e. a Windows version of the program. However there is also a
console version ( SciLex.exe ) of the program.
The console version can be driven by VBS via SendKeys to issue instructions to
build the c:\Temp\Data6.TXT file. This file in turn can be used along with a
shell of your FIR AFL to build AFL for whatever Cycle Period and in turn
whatever Filter Length ( or should it be the other way around ? ) is desired.
I've been playing with a .VBS to perform this function this morning and while
still a little rough around the edges it seems to do the job.
Here's the VBS ( Keep in mind it is for SciLab v4.0 ) ...
At the moment it is set up to:
- Run SciLab ( SciLex.exe ) to calculate the Coefficients for a 400 Period Cycle
and write those to c:\Temp\Data6.TXT
- Read the c:\Temp\Data6.TXT file that is created by SciLab and create a
FIR_FL_CP.AFL ( at the moment also in c:\Temp ) where FL = FilterLength and CP =
CyclePeriod. At the moment the Cycle Period is hard wired at 400 but it would be
simple enough to have this create multiple AFL's with whatever CyclePeriod is
desired.
Here is the output ( generated ) AFL which I have not played with at all except
to test that it was synyax free in AB / AA.
The next logical place to
go with this would seem to be to have an automated way of perfmorning the manual
functions you do now with the various param statements in the AFL.
Fir.zip 8/1/06 Fred Tonetti-Revised version here ...