Cycle Analysis

If you don't know where to start, start here! Don't be afraid to ask questions.

Moderator: moderators

acepsut
rank: <50 posts
rank: <50 posts
Posts: 6
Joined: Wed Aug 29, 2007 8:52 am
Reputation: 0
Gender: Male

Cycle Analysis

Postby acepsut » Thu Jan 15, 2009 4:16 pm

[font=Comic Sans MS]Good morning,

I am new and this is my first post.

With the help of a friend we've developed a cyclical approach to model and predict market behavior, from a scientific point of view.

We got very nice results, and we think we can get even more with an in-depth study and analysis of every step of the whole process.

For this reason we are seeking for people with good background in mathematics to create a study group about cycles and periodicities in financial markets; engineers, mathematicians, physicists etc. are all welcome.

The attach is a sample of what has been done so far, of course if we always get such results we do not need to study anymore...

If there is somebody interested in, please contact me directly via Skype, my nick is acepsut

Greetings from Italy and apologize for my bad english.

Alberto[/font]
Attachments
A730.png
dax 60 min predict + out of sample verify
A730.png (17.51 KiB) Viewed 335 times

Please add www.kreslik.com to your ad blocker white list.
Thank you for your support.

Tresor
rank: 50+ posts
rank: 50+ posts
Posts: 93
Joined: Sat Mar 15, 2008 2:30 pm
Reputation: 0
Location: Davy Jones' Locker
Gender: Male

Re: Cycle Analysis

Postby Tresor » Fri Jan 16, 2009 3:04 pm

acepsut wrote:[font=Comic Sans MS]Good morning,

I am new and this is my first post.

With the help of a friend we've developed a cyclical approach to model and predict market behavior, from a scientific point of view.

We got very nice results, and we think we can get even more with an in-depth study and analysis of every step of the whole process.

For this reason we are seeking for people with good background in mathematics to create a study group about cycles and periodicities in financial markets; engineers, mathematicians, physicists etc. are all welcome.

The attach is a sample of what has been done so far, of course if we always get such results we do not need to study anymore...

If there is somebody interested in, please contact me directly via Skype, my nick is acepsut

Greetings from Italy and apologize for my bad english.

Alberto[/font]


Your work seems to be similar to Clyde Lee's studies on cycles. Try him at theswingmachine.com

Regards
Attachments
junk.gif
junk.gif (49.52 KiB) Viewed 326 times

acepsut
rank: <50 posts
rank: <50 posts
Posts: 6
Joined: Wed Aug 29, 2007 8:52 am
Reputation: 0
Gender: Male

Postby acepsut » Fri Jan 16, 2009 3:45 pm

Hello Tresor,

I know Clyde Lee's work but our is quite different.

Htarlov
rank: <50 posts
rank: <50 posts
Posts: 13
Joined: Mon Dec 15, 2008 8:30 pm
Reputation: 0
Gender: None specified

Postby Htarlov » Sat Jan 17, 2009 10:00 pm

Did you tried your study against statistically significant amount of samples?

We were trying to use moving fourier transform with some long window on EURUSD, USDJPY, DAX, CAC and GC time series, then use some part frequencies spectrum and inverse fourier transform to make a model for prediction of future values. But for now without statistically significant result.

acepsut
rank: <50 posts
rank: <50 posts
Posts: 6
Joined: Wed Aug 29, 2007 8:52 am
Reputation: 0
Gender: Male

Postby acepsut » Sun Jan 18, 2009 11:10 am

Hi Htarlov,

we do not have statistical results since we know there is a lot to do to improve the whole process we've done so far.

We only see very nice results with very accurate tracking of out of sample data, and sometimes completely wrong.

I think there are times we can expect good modeling and prediction depending on market caractheristics itself, and times when this study will not be able to.

In example, if our serie is made of only noise we cannot use this approach and it is better do not trade on periodicities.

I do not think cycle analysis is to be used every time and in every market conditions, but rather to find out when to rely on higher probability of good market modeling based on cycles.

For this reason I would like to join the efforts to create a group for a scientific approach to cycle analysis, to study and apply original ideas to what has been proposed in this field.

We are available to discuss, compare and, of course, work togheter.

Regards,

Alberto

Please add www.kreslik.com to your ad blocker white list.
Thank you for your support.

obx
rank: <50 posts
rank: <50 posts
Posts: 41
Joined: Fri May 26, 2006 12:35 pm
Reputation: 0
Gender: Male

Postby obx » Tue Jan 20, 2009 9:10 pm

Alberto,

Please organize a DmDm meeting with some invitees to demonstrate and discuss what you are doing and what you would like to accomplish with cycles. Link is DmDm. On DmDm up to 4 participants can be 'miked' at a time on the free version and you can post images or live screens to illustrate what you're talking about.
If invited, I'll be there...
Many thanks

obx

acepsut
rank: <50 posts
rank: <50 posts
Posts: 6
Joined: Wed Aug 29, 2007 8:52 am
Reputation: 0
Gender: Male

Postby acepsut » Wed Jan 21, 2009 9:41 am

Hello obx,

I never heard about DmDm but no problem I can organize a conference if we need to.

What I would like to say is that neither I posted here to let all know how good we are nor to sell anything. Simply we've developed a cycle based work according to what Herbst suggest in his book but with original ideas and variations.

Althoug the code was not developed in full details and analysis, we often get surprised in how accurate out of sample prediction match real moves, and follow closely for many points ahead.

This is not a game of up or down, or 50% possibility since the output is a composite of many cycles and plot prediction with many turning points, slopes and different directions.

Just to answer to a few common questions may one ask:

1) Does your work track perfectly every time?

No, and I think it's impossible to have always accurate predictions, rather I prefer to think to rely in moments of higher probabilities of success.

2) Are you seeking for someone to sell your code?

No, absolutely no!

3) What's the reason for your post here in Kreslik forum?

I want to gather people with good maths and/or dsp background, I strongly believe that good and reliable results can be achieved from a team job. Works done so far has been developed in Matlab and we prefer to continue in this environment, maybe later we can think to a stand alone platform.

Once again, we think cycle analysis can lead to really accurate results if the job is done in a scientific approach and with an in-depth study of every step.

My request will be spread worldwide in other trading forum.

For now, anyone interested in collaborate can add me on skype, my nick name is acepsut

Regards,

Alberto

obx wrote:Alberto,

Please organize a DmDm meeting with some invitees to demonstrate and discuss what you are doing and what you would like to accomplish with cycles. Link is DmDm. On DmDm up to 4 participants can be 'miked' at a time on the free version and you can post images or live screens to illustrate what you're talking about.
If invited, I'll be there...
Many thanks

obx

obx
rank: <50 posts
rank: <50 posts
Posts: 41
Joined: Fri May 26, 2006 12:35 pm
Reputation: 0
Gender: Male

Postby obx » Thu Jan 22, 2009 8:41 pm

Alberto,

Do you use amplitude of individual cycles in your work?

What time of day is best to skype you?

thx

obx

acepsut
rank: <50 posts
rank: <50 posts
Posts: 6
Joined: Wed Aug 29, 2007 8:52 am
Reputation: 0
Gender: Male

Postby acepsut » Fri Jan 23, 2009 7:29 am

Hello obx,

we haven't done any study about amplitude modulation and this is one of the task we'd like to work on.

My time zone setting is GMT+1 and you can find me from 9 am to 6 pm.

Regards,

Alberto

obx wrote:Alberto,

Do you use amplitude of individual cycles in your work?

What time of day is best to skype you?

thx

obx

acepsut
rank: <50 posts
rank: <50 posts
Posts: 6
Joined: Wed Aug 29, 2007 8:52 am
Reputation: 0
Gender: Male

Cycle analysis on eur/usd

Postby acepsut » Fri Jan 23, 2009 9:42 am

[font=Comic Sans MS]Just a quick look at how cycle code behave in different market, this is a sample from eur/usd 15 minutes time frame.

As you can see, there is no amplitude modulation but I think quality predict speaks for itself.

As a reminder, the "in sample" data (known data) ends at point 60, and blue line is forecast. Red bars are later overplotted to verify forecast accuracy.[/font]
Attachments
A877.png
A877.png (12.28 KiB) Viewed 248 times

Please add www.kreslik.com to your ad blocker white list.
Thank you for your support.


Return to “beginners forum”