That's why de-trending and windowing must be applied before doing FFT.

Best regards,
Tomasz Janeczko
amibroker.com
----- Original Message ----- 
From: "Fred" <[EMAIL PROTECTED]>
To: <[email protected]>
Sent: Wednesday, October 11, 2006 5:33 AM
Subject: [amibroker] Re: Hurst Channels Code


> As I recall Hurst didn't make much more mention of it in the course 
> material then in PM ... However for those wanting to understand his 
> methodologies the course is the definitive work which goes well 
> beyond CMA's & DE's ... DE's can be an interesting tool for trading 
> at the extremes ... I'll post a version later that uses PolyFit for 
> extrapolation and has a variety of flexibility to it not present in 
> the AFL that's kicking around at the moment.  My personal preference 
> is for Trig Fits as opposed to a polynomial Fit, but when properly 
> damped the PolyFits work ok too.
> 
> IMHO the periodogram is nothing more then a product of an FFT like 
> algorithm i.e. an investigative tool for determining cyclic 
> forces ... There are a variety of methodologies and tools out there 
> to produce one and from what I understand at least one will be built 
> into AB on the next release.  Personally I have several problems with 
> FFT's ... which can be demonstrated with a wave generator ... 
> 
> For example the code below ( It's a VBS out take ... Not complete, 
> but usable with minor mods ) will generate a composite wave 
> consisting of several cosines of various user selected wave lengths, 
> amplitudes & phase offsets ... it can also add a trend to the 
> composite as well as random noise to almost any level ... It is IMHO 
> very instructive to use stuff like this and see how they affect 
> different DSP algorithms.
> 
> FFT's have a tendancy to be easily fooled by trend and even more so 
> by a combination of trend and noise.  This leads one to surmise that 
> it is probably a good idea to at least remove the trend from data 
> before tossing it to an FFT for analysis.  The other problems that 
> FFT's have center around their inability to have relatively good 
> resolution at low frequency.  The reason of course is related to how 
> the algorithm itself works.  This sort of forces one to use long data 
> histories relative to the cycle lengths being investigated so that 
> the result is decent resolution.  The problem with this approach is 
> that IMHO wave lengths either change or if you are a purist some 
> times are overwhelmed by non cyclic activity i.e. 10/87 etc ... Since 
> FFT's in essence provide data related to cycle length, amplitude and 
> phase offset FROM THE BEGINNING OF THE DATA, the shorter wave 
> lengths, which are also typically buried in noise, don't extrapolate 
> correctly.  So while FFT's can definitely provide some insigth into 
> the cyclic forces they are not a be all and end all.  
> 
>    pi = 4 * atn(1)
> 
>    Randomize
> 
>    StartX = (dFactor + 1) * 1000
> 
>    for i = 1 To nIS + nCoeff
> 
>        Trend(i) = (i - 1) * rTrend * dFactor * (rCyc1Amp + rCyc2Amp 
> + rCyc3Amp + rCyc4Amp + rCyc5Amp) / 5
>        Noise(i) = (Rnd - 0.5) * 2 * rNoise * dFactor * (rCyc1Amp + 
> rCyc2Amp + rCyc3Amp + rCyc4Amp + rCyc5Amp) / 5
> 
>        Cycle1(i) = cos(((i - 1) / rCyc1Len + rCyc1Phase / 360) * 2 * 
> pi) * (rCyc1Amp * dFactor) + Trend(i) * 0.2
>        Cycle2(i) = cos(((i - 1) / rCyc2Len + rCyc2Phase / 360) * 2 * 
> pi) * (rCyc2Amp * dFactor) + Trend(i) * 0.2
>        Cycle3(i) = cos(((i - 1) / rCyc3Len + rCyc3Phase / 360) * 2 * 
> pi) * (rCyc3Amp * dFactor) + Trend(i) * 0.2
>        Cycle4(i) = cos(((i - 1) / rCyc4Len + rCyc4Phase / 360) * 2 * 
> pi) * (rCyc4Amp * dFactor) + Trend(i) * 0.2
>        Cycle5(i) = cos(((i - 1) / rCyc5Len + rCyc5Phase / 360) * 2 * 
> pi) * (rCyc5Amp * dFactor) + Trend(i) * 0.2
> 
>        CycleX(i) = Cycle1(i) + Cycle2(i) + Cycle3(i) + Cycle4(i) + 
> Cycle5(i) + Noise(i)
> 
>        ISData(i) = StartX + CycleX(i)
> 
>    Next
> 
> 
> --- In [email protected], "wavemechanic" <[EMAIL PROTECTED]> wrote:
>>
>> Fred:
>> 
>> Did you go through the course 
> (http://www.traderspress.org/detail.asp?product_id=1400)?  Do you 
> understand how the Periodogram is calculated?  That's one of the big 
> blanks in my understanding of the Hurst world.  Based on the 
> TraderPress summary that seems to be covered in the course.
>> 
>> Bill
>> 
>> 
>> ----- Original Message ----- 
>> From: "Fred" <[EMAIL PROTECTED]>
>> To: <[email protected]>
>> Sent: Tuesday, October 10, 2006 9:23 AM
>> Subject: [amibroker] Re: Hurst Channels Code
>> 
>> 
>> > Most of us who have read Hurst's writtings are well aware of the 
>> > general principles involved in PM ... However, there are several 
>> > points that should be mentioned or questioned with regards to 
> your 
>> > comments ...
>> > 
>> > 1. It is clear from Hurst's course material that PM and the 
>> > techniques therein were, in his mind, a fairly simplistic view of 
> how 
>> > he analyzed price and that the curvelinear trendlines were only a 
>> > means for the analyst to discover the cyclic forces in play i.e. 
> a 
>> > possible first step in developing a full phasing Analysis ( FPA ) 
> and 
>> > using FLD's, VTL's and other techniques ...
>> > 2. As far as the tools and the math involved in the later 
> chapters 
>> > and appendicies it is also clear that Hurst not only employed the 
>> > techniques therein but other more sophisticated techniques that 
> he 
>> > didn't really write much about in PM.  Clues to this can be seen 
> in 
>> > the bibliography.
>> > 3. While PC's and AB didn't exist in 1970, mainframes and mini's 
> did 
>> > and if you recall we had already managed to get to the moon and 
> back 
>> > mostly by use of the slide rule.
>> > 4. The curvelinear bands that Hurst's uses are in essence n 
> period 
>> > centered moving averages with percentage or standard deviation 
> bands 
>> > around them which by their nature can not be computed closer than 
> n / 
>> > 2 bars prior to the current bar.  As such one must have some 
>> > technique for extrapolating those bands to at least the current 
> bar 
>> > and preferably beyond.  Without this capability one would have 
>> > already experienced half the move in the opposite direction when 
> the 
>> > CMA catches up.
>> > 
>> > My question then is ... If you are using curvelinear bands of one 
>> > sort or another inside AB how are you constructing the bands and 
> what 
>> > methodologies are you using to extrapolate them ?
>> > 
>> > --- In [email protected], "Ton Sieverding" 
>> > <ton.sieverding@> wrote:
>> >>
>> >> Hurst wrote 'The Profit Magic of Stock Transaction Timing' in 
> 1970. 
>> > When the technical annalist was using pencil and paper for his 
>> > graphs. Bill Gates and TJ still did not exist and Hurst was not 
> using 
>> > a PC. I've tried to get the original version of the book but only 
> got 
>> > the reprinted version from 2000. Frankly I have the feeling that 
>> > Chapter 11 as well as the Appendices have been added at the time 
> of 
>> > the reprint. This to give you my opinion how I see the book. It 
> is an 
>> > attempt to work with forecasted envelopes. And a good one ...
>> >> 
>> >> For me the practical 'red line' for using the Hurst principle is 
> as 
>> > I told you already rather simple. Try to construct two trading 
> bands. 
>> > A long and a short period trading band. Hurst tells you how to do 
>> > that with pencil and paper. The software today does it 
> automatically 
>> > for you. As soon as the short band is above the long one and the 
>> > price above the short band you will have an 'overbought' 
> situation. 
>> > When the short band falls below the long band and the price is 
> below 
>> > the short band you have an 'oversold' situation. The combination 
> of 
>> > these 'overbought' and 'oversold' situations with a falling resp. 
>> > rising long trend will give you the SELL and BUY signals. I am 
> using 
>> > Hurst with weekly views and indexes and must say that the results 
> are 
>> > very good ...
>> >> 
>> >> Ton.
>> >> 
>> >>   ----- Original Message ----- 
>> >>   From: Rakesh Sahgal 
>> >>   To: [email protected] 
>> >>   Sent: Monday, October 09, 2006 3:18 PM
>> >>   Subject: Re: [amibroker] Hurst Channels Code
>> >> 
>> >> 
>> >>   I suggest you get familiar with the concept of cycles a la 
> Hurst 
>> > and
>> >>   use of Curvilinear Bands before you try to comprehend the 
> code. As
>> >>   for the code looking into the future - yes it does.
>> >> 
>> >>   R
>> >> 
>> >>   On 10/9/06, Saji Oommen <sajioommen2@> wrote:
>> >>   > Hello,
>> >>   >
>> >>   > Thanks for sharing the code. Could you kindly confirm
>> >>   > that does this code have any ZIG function in it or
>> >>   > does it look into future or the signals change when
>> >>   > new data comes in. I read through the code and could
>> >>   > not make it out. I am in the learning process of AFL.
>> >>   > So pardon my ignorance.
>> >>   >
>> >>   > Regards
>> >>   >
>> >>   > Saji
>> >>   >
>> >>   > --- Ton Sieverding <ton.sieverding@> wrote:
>> >>   >
>> >>   > > The way I see it is rather simple. Let's take the
>> >>   > > S&P500. Using a weekly view :
>> >>   > > a.. Buy when Blue+Red+Price BELOW Green and Green
>> >>   > > rising channel
>> >>   > > b.. Sell when Blue+Red+Price ABOVE Green and Green
>> >>   > > falling channel
>> >>   > > Sell Augustus 2000,
>> >>   > > Buy March 2003,
>> >>   > > ... still Long.
>> >>   > >
>> >>   > > Ton.
>> >>   > >
>> >>   > > ----- Original Message -----
>> >>   > > From: Rakesh Sahgal
>> >>   > > To: Amibroker Yahoogroup
>> >>   > > Sent: Monday, October 09, 2006 10:57 AM
>> >>   > > Subject: [amibroker] Hurst Channels Code
>> >>   > >
>> >>   > >
>> >>   > > For any of those interested in tinkering with
>> >>   > > Hurst Channels, am
>> >>   > > attaching code I had recieved from group member
>> >>   > > Jason Hart. This code
>> >>   > > was written by group member going by the nickname
>> >>   > > "Ace".
>> >>   > >
>> >>   > > I was not able to make much use of it. The reason
>> >>   > > was - the values of
>> >>   > > the channels, current and in the immediate past
>> >>   > > i.e. the projection
>> >>   > > zone, were not constant. Since the values of the
>> >>   > > projections changed
>> >>   > > everyday I had no way of making any meaningful use
>> >>   > > of it ( I was
>> >>   > > basically looking for what Hurst called "Edge
>> >>   > > Band" Trades). If my
>> >>   > > understanding was incorrect and anyone can make
>> >>   > > sense of them, please
>> >>   > > be kind enough to share your insights with the
>> >>   > > rest of us.
>> >>   > >
>> >>   > > Rakesh
>> >>   > >
>> >>   > >
>> >>   > >
>> >>   > >
>> >>   >
>> >>   >
>> >>   > __________________________________________________
>> >>   > Do You Yahoo!?
>> >>   > Tired of spam? Yahoo! Mail has the best spam protection 
> around
>> >>   > http://mail.yahoo.com
>> >>   >
>> >>   >
>> >>   > Please note that this group is for discussion between users 
>> > only.
>> >>   >
>> >>   > To get support from AmiBroker please send an e-mail directly 
> to
>> >>   > SUPPORT {at} amibroker.com
>> >>   >
>> >>   > For other support material please check also:
>> >>   > http://www.amibroker.com/support.html
>> >>   >
>> >>   >
>> >>   > Yahoo! Groups Links
>> >>   >
>> >>   >
>> >>   >
>> >>   >
>> >>   >
>> >>   >
>> >>   >
>> >>   >
>> >>   >
>> >>   >
>> >>
>> > 
>> > 
>> > 
>> > 
>> > 
>> > 
>> > Please note that this group is for discussion between users only.
>> > 
>> > To get support from AmiBroker please send an e-mail directly to 
>> > SUPPORT {at} amibroker.com
>> > 
>> > For other support material please check also:
>> > http://www.amibroker.com/support.html
>> > 
>> > 
>> > Yahoo! Groups Links
>> > 
>> > 
>> > 
>> > 
>> > 
>> > 
>> > 
>> > 
>> > 
>> > 
>> > 
>> > 
>> > 
>> > -- 
>> > No virus found in this incoming message.
>> > Checked by AVG Free Edition.
>> > Version: 7.1.407 / Virus Database: 268.13.1/466 - Release Date: 
> 10/7/2006
>> > 
>> >
>>
> 
> 
> 
> 
> 
> 
> Please note that this group is for discussion between users only.
> 
> To get support from AmiBroker please send an e-mail directly to 
> SUPPORT {at} amibroker.com
> 
> For other support material please check also:
> http://www.amibroker.com/support.html
> 
> 
> Yahoo! Groups Links
> 
> 
> 
> 
> 
> 
> 
> 
> 
>


Please note that this group is for discussion between users only.

To get support from AmiBroker please send an e-mail directly to 
SUPPORT {at} amibroker.com

For other support material please check also:
http://www.amibroker.com/support.html

 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/amibroker/

<*> Your email settings:
    Individual Email | Traditional

<*> To change settings online go to:
    http://groups.yahoo.com/group/amibroker/join
    (Yahoo! ID required)

<*> To change settings via email:
    mailto:[EMAIL PROTECTED] 
    mailto:[EMAIL PROTECTED]

<*> To unsubscribe from this group, send an email to:
    [EMAIL PROTECTED]

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/
 



Reply via email to