Fast Fourier Transform - Cycle Extraction - page 33

 

debashis,

Take your time. You will have to understand how to use Rapidminer before you will be able to do anything useful with it. It would be the same with MATLAB or R or any of the tools available. The learning curve is steep and I doubt you will be able to get far in a day.The wavelets on their own will not help you with your trading until you know how to use them. I posted a script that was far from complete so that there would be an example here of what MODWT looks like. People with experience looking at cycle curves would immediately know what they were looking at with no surprises.

Look at Maximum Entropy Spectrum Analysis, look at Goerzel, by all means look at wavelets but from a standpoint of understanding what the differences are. Read the papers for a start! Only then can you decide if these tools can make a difference to your style of trading. These tools are seductive because sometimes they are so close yet often so far away from ideal when it comes to real world trading that we might as well be looking at the moon.

There is a thread in the Advanced Elite section with an exporter that outputs csv files that are specific to Rapidminer. I am sure there are a few general ones in various places as well.

Don't give up though, this is a fascinating topic that will keep you busy for a long time!

 

Hurst

The following gif shows the Nominal cycle model of J.M Hurst:

ImageShack� - Online Photo and Video Hosting

You can always compare it to the actual cycle lengths by using the "Composite" indicator of post #288. Many thanks to mladen for this interesting solution.

Sixer

 

Thanks Sixer for the Hurst cycle sheet its a good reference so I'll attach it to my post

I also added how the harmonics of the cycles looks like " credit to http://www.decisionpoint.com/tacourse/Cycles3.html"

So with regards to delta points or delta phenomena , Most of these points will correspond and match the major troughs from Hurst cycles , they are also the turing points on the Fourier solution , they are the same points you would use to cluster the signal for wavelet analysis, they will match the Elite wave points and might match gann squares as well.

I am saying MOST as each of the trading methods mentioned above use different approach to slice the market , but at the end the solution is MOSTLY common.

Edit : I added another screen shot from sentienttrader.com

A live trading example would be the pairs (xxx.xxx) I am currently shot on this pair ( I scale the profit as you can see from the order number ) and I am expecting this pair to drop significantly within the next 10 days , after that the short term cycle would end and regardless I would have to close that trade , so the trade is mainly defined by time rather than price.

 
MiniMe:
There is an related topic on FF by the title " Optimized Trend Trading".I found page 75 and 76 quite interesting.

Specially what "MadCow" wrote in his post , he added a link to his previous thesishttp://onlinesvr.altervista.org/ the site is informative and have some codes.

BTW Support Vector Machine SVM is the advanced approach of wavelet analysis, if you want to compare wavelet with SVM its like a vector to a matrix.

This is only a guess not supported by data:This can be an another interesting and suitable SVM algo : support vector echo-state machines (SVESMs)

'' The basic idea is replacing "kernel trick" with "reservoir trick" in dealing with nonlinearity, that is, performing linear support vector regression (SVR) in the high-dimension "reservoir" state space'' Software - Zhiwei SHI 's Homepages!

 

MTF Version

mladen:
Pip

This is a chop-chop job (when you take a look at the code , you will understand thy I say "chop-chop" ) done on the indicator in order to allow start bar different than 1 (which was the problem with previous versions from the start)

As far as I tried it, it is working OK (again the biggest problem was the "future" : determining the exact time of the future bar when, for example there is a gap between Friday and Monday) Anyway, it seems to be working OK now. Here is an example with a 25 as a start bar both for "simple" and "composite"

Mladen, is it possible to make a MTF version of this indicator ? I'd like to be able to set an option, which would be to use original timeframe data, but showing it on another timeframe. For example showing lower timeframe data on higher ones or higher ones on lower ones. So for example looking at M30, I'd like to put an indi showing M5 cycles in real time action. So the indi would not be related to time anymore, it still uses original timeframe data, but just showing on another timeframe. I hope it's understandable and possible

 

Echo MiniMe's thoughts

Dear members,

Here is a quick screenshot that adds support to cycle application in trading as well as Minime's thoughts.

In the attached image i have used the Specter tool with modified settings to align the cycle properly to price action and apply only most applicable cycles. As you can see, it did a pretty good job in predicting market movement for today.

Naturally, i used supporting evidence (e.g. PMP and TMP) to confirm trades. As of the time of writing this post I am long EU at 2870 and looking to exit 32 pips later.

The settings adjustment is performed manually at the moment until am able to determine the algorithm needed to automate it. But one thing I am certain will need to be done is

1) Separate the cycles so that we can control the start of each cycle separately.

2) Control the phase of each cycle to match it with market action

The "how" i haven't figured out yet

Cheers,

Pip

Files:
 

...

yeah....right Mr. Pip...prevailing cycle must prewhale:)

 
Pip:
Dear members,

Here is a quick screenshot that adds support to cycle application in trading as well as Minime's thoughts.

In the attached image i have used the Specter tool with modified settings to align the cycle properly to price action and apply only most applicable cycles. As you can see, it did a pretty good job in predicting market movement for today.

Naturally, i used supporting evidence (e.g. PMP and TMP) to confirm trades. As of the time of writing this post I am long EU at 2870 and looking to exit 32 pips later.

The settings adjustment is performed manually at the moment until am able to determine the algorithm needed to automate it. But one thing I am certain will need to be done is

1) Separate the cycles so that we can control the start of each cycle separately.

2) Control the phase of each cycle to match it with market action

The "how" i haven't figured out yet

Cheers,

Pip

Maybe I haven't understood your post, but can't you just put one indicator for each cycle so you can set each of them differently ?

 
airquest:
Maybe I haven't understood your post, but can't you just put one indicator for each cycle so you can set each of them differently ?

true, except i want all to be present in the same window to generate a composite off, plus it would take too much screen real estate

 

outcome of trade, closed for break-even plus 3 pips.

Cheers,

Pip

Pip:
Dear members,

Here is a quick screenshot that adds support to cycle application in trading as well as Minime's thoughts.

In the attached image i have used the Specter tool with modified settings to align the cycle properly to price action and apply only most applicable cycles. As you can see, it did a pretty good job in predicting market movement for today.

Naturally, i used supporting evidence (e.g. PMP and TMP) to confirm trades. As of the time of writing this post I am long EU at 2870 and looking to exit 32 pips later.

The settings adjustment is performed manually at the moment until am able to determine the algorithm needed to automate it. But one thing I am certain will need to be done is

1) Separate the cycles so that we can control the start of each cycle separately.

2) Control the phase of each cycle to match it with market action

The "how" i haven't figured out yet

Cheers,

Pip
Files:
Reason: