Forex Factory (https://www.forexfactory.com/forum.php)
-   Trading Discussion (https://www.forexfactory.com/forumdisplay.php?f=11)
-   -   Optimized Trend Trading (https://www.forexfactory.com/showthread.php?t=68181)

fajst_k Nov 13, 2010 6:19am | Post# 841

Repainting/Recalculation is a phenomenon. Why it repaints. Because it uses future data for the calculations. To minimize this fact I use a lag, and that is the simplest but not the optimum solution. another approach is to tune the SSA to the market conditions or to use a neural net (or something else) to predict the future bars and make a calculation based on that, but if you can predict the future bars why should you use SSA LOL/There are guys out there who claim the whole approach of SSA is proven useless and Independent Component Analysis or...
This is how NOXA CSSA is working using echo state NN. See thread on TRAD2WIN - i checked the reality of this

fajst_k Nov 13, 2010 6:24am | Post# 842

Look at it Yourself. In fact I am not sure even if it is a SSA algorithm.
Install the both SSA files into the library folder.
I know those indicators. The question is how exactly you lagged the data to SSA to suppress the future leak ?

John Last Nov 13, 2010 6:34am | Post# 843

1 Attachment(s)
To tell the truth I know what happens and how they recalculate, and that was not the idea at all. I wanted to show the fractal break - out pattern.

But I will tell you exactly what I did. I put this indicator together with the PFE indicator in the same window. By the drag and drop. The fact that some guy remarked that surprised me because I did nothing at all and I even did not see that.

Here is the other indicator I add in the same window.

When the markets are open I will closely examine it on 1 m time frame, because likely this is a mirage. But I want to understand what happened.

And maybe the problem is more profound. If we see a recalculating indicator not repainting we can ask why. Probably it is related with the market conditions and that upon some circumstances the SSA does not repaint but predicts ?! Is there a pattern?

As far as I know all the oscillating patters have a chaotic nature. They are chaotic attractors and maybe part of some more complex patterns.
My opinion is that the SSA describe and we need something else to look behind. Nevertheless my opinion is more a chaotic attractor is complex less stable it is. The point is how to use the simplest form of attractor.
PFE.mq4

fajst_k Nov 13, 2010 7:09am | Post# 844

1 Attachment(s)
To tell the truth I know what happens and how they recalculate, and that was not the idea at all. I wanted to show the fractal break - out pattern.

But I will tell you exactly what I did. I put this indicator together with the PFE indicator in the same window. By the drag and drop. The fact that some guy remarked that surprised me because I did nothing at all and I even did not see that.

Here is the other indicator I add in the same window.

When the markets are open I will closely examine it on 1 m time frame, because likely this is a mirage....
So the results from post 838 were obtained automatically or manually ??

Anyway rule is quite simple. If you have just one element which has a future leak in your system results will be false - too good. So perhaps the best to
desk check all indicators used if they don't have future leak coding.

Attached help of NOXA CSSA, chapter CSSA 101 can be helpfull
NoxaCSSA.zip

John Last Nov 13, 2010 8:06am | Post# 845

1 Attachment(s)
Hi what about computing the lag of the SSA based on a half of a dominant cycle by a DSP method?

And if we have a good signal to noise ratio?

And good continuous dominant cycle period?

Negative conditions: absence of a shift in the fractal structure.

A signal would be when it crosses another oscilator, the Polarysed Fractal Efficiency in this case.

Look at the cycle period, nice continuous cycle period. Even if it is lagging it is OK.

On this shot you can see the Corona cylce period.
Click to Enlarge

Name: ex3.gif
Size: 44 KB

fajst_k Nov 13, 2010 9:12am | Post# 846

Hi what about computing the lag of the SSA based on a half of a dominant cycle by a DSP method?

And if we have a good signal to noise ratio?

And good continuous dominant cycle period?

Negative conditions: absence of a shift in the fractal structure.

A signal would be when it crosses another oscilator, the Polarysed Fractal Efficiency in this case.

Look at the cycle period, nice continuous cycle period. Even if it is lagging it is OK.

On this shot you can see the Corona cylce period.
What dominant cycle ?? Do you believe in this ?? Put random time series and
Corona will find DC also. Those cycles lasting too short to be dedected, some people claim to dedect them on high TF or daily TF but as usual there is never enough OOS trades to make any conclusions.

I posted on tsd digital filters thread some test signals in MT4 format. Just apply corona to this and you will see.

make strategy according to your proposal, fully automatic without repainting indicators and you will see results

John Last Nov 13, 2010 11:06am | Post# 847

Determinism test
 
Before making any assumptions. What about a determinism test. It is important to know what is going on the market now.

We have to verify that the studied time series originate from a deterministic system.

We have to distinguish deterministic chaos from irregular random behavior that resembles to chaos.

If a system is deterministic it can be explained by a set of differential equations. So the vector field in any point of the phase space is determined by the differential equations.

It is like black jack game and we have to count the cards in the casino once we see that only the high cards are remaining we can make big bets.

Here it is the same thing. We have to determine a market state that is deterministic, so we can play. If we are in the right moment e.g. all the Neural Nets will work, and they won't otherwise.

In practice we are trying to reconstruct a cycle chaotic attractor from the time series. And the attractor is in multi dimensional phase space. Nice Ah.
All the oscilators are just trying to reconstruct a cycle chaotic attractor by inappropriate ways. Is it possible to make any better? The DSP approach of Ehlers is using a MESA for very short periods. This is nonsense from a DSP perspective but a sound logic from a chaotic perspective (because as normally the maximum Lyapunov exponent is positive the system is loosing memory with the time and becomes unpredictable). So even if the cycle measure is useless it makes no harm.

mbkennel Nov 14, 2010 8:19pm | Post# 848

Before making any assumptions. What about a determinism test. It is important to know what is going on the market now.

We have to verify that the studied time series originate from a deterministic system.
There is no stationary low dimensional determinism in financial markets in the classical ODE sense, but that doesn't mean all hope is lost.

John Last Nov 15, 2010 4:08am | Post# 849

Yep there is, there is.
And when we have it something strange happens. I call this a low phase space singularity.

In this case the phase space of the possible solutions is really low. In that phase space a lot of algorithms are able to find a solution simultaneously. And they find and act on that solution simultaneously.

From a game theory perspective the humans have in principle non cooperative behavior. The machines have a cooperative behavior, very different algorithms are able to find the exact same solution in the exact same time. So now we are in the Isaac Asimov time, we have not only to consider the human psychology but the Robot psychology too LOL.

The problem is that those algorithms operate not in the human time but in their own time of the high frequency and microseconds. When the singularity occurs they start to drive the price, acting in synchronization. That allows us humans to see the action in our time of reference.

When that low phase space singularity happens all prediction becomes in fact impossible at our time level of reference and we have just to follow the movement, with tuned digital filters. I call it Persistent Movement Following (PMF) Strategy, a.k.a Black Noise Hunter. The movement is persistent and the black noise phenomena dominate (e.g. the price drives in one direction after that it changes a direction and drives again).

As an observation every actual crash is related with this kind of phenomenon. I usually detect the beginning stage of the singularity by a peak of the Hurst difference indicator. When we have a strange peak, something is going on. The peak is revealing that there is a very fast change inf the fractal characteristics of the price.

John Last Nov 15, 2010 7:24am | Post# 850

Repainting or not?
 
1 Attachment(s)
Hi,

Here I just add a screen shot of the SSA Normalize indicator.
Is it repainting or it is not repainting. If we have calculated well the lag period it should not be repainting. If it repaints that reveals an important information.

So that it why we have arrows, it indicates what was indicated by the indicator in case it repaints in that way we have a track.

I find it to be the best oscillator, combined with another oscillator of choice. So the oscillator is giving indication what is going on, and the SSA is giving indication what should happen. If the normal oscillator goes in the direction indicated by the SSA that is an information. If we are going to use that information or not it is a matter of strategy.

So as far I understand with a SSA we are trying to keep a track of the Cycle chaotic attractors. The fact that the SSA uses future data is not a problem, we lag the SSA. So the information even if it is lagged concerns the present moment. And the present and the past moments have an information about the near future.

I have to find better for information about ranging conditions. You can change the PFE with a digital filter of your choice.

Remeber a SSA cannot predict a break - out.

Yes there is a repainting it is very subtle

I detect a change in the spectrum so according to the tentative theory I will add more lag because I want to track the highest order of chaotic attractor. According to the tentative theory we do not have cycles we have chaotic attractors living in multidimensional phase space. Tracking cycles with DSP cycles identification techniques can give rough hint where those cycles could be found and what is their activity. This is a theory.
There is a joke that for the hammer everything is a nail. LOL LOL LOL.
Click to Enlarge

Name: repaint.gif
Size: 46 KB

michaelwen Nov 22, 2010 3:56am | Post# 851

MR,John Last: very good information you provid in here , keep up!!! great work!

John Last Nov 23, 2010 11:14am | Post# 852

Low Phase singularity examples
 
3 Attachment(s)
Hi,

Thank you for the comments. I will just try to give an example how it looks like in a screen shot.

I think that if you have eyes you will see a pattern. To make things easier I have put some comments on the graph.

In fact this analysis is totally independent from the Technical Analysis. The good point is that it combines very well with it, because it is an independent form of analysis.

I have added a template worth digital filters and what they give, but the idea is so simple that you do not even need it as it is seen on the naked screenshot.

You can see several patterns:

Fractal break - out. Quick transition from blue to red, high fractal dimension towards lower fractal dimension.

This fractal break - out may transform into a low phase space singularity. Normally the FGDI gets lower and lower.
When it flattens you can expect a Black noise phenomenon occur.

If that happens... You see.
Click to Enlarge

Name: bt15.gif
Size: 30 KB
Click to Enlarge

Name: bt30.gif
Size: 30 KB
Click to Enlarge

Name: singularity.gif
Size: 25 KB

John Last Nov 24, 2010 4:11am | Post# 853

Non Repainting SSA
 
1 Attachment(s)
The guys from TSD have made a great breakthroughs in the use of SSA.
In fact what I have done previously is to add a digital smoothing of the SSA, so it would not react to a minor change and would react at an important change and in the same time to use 20 bars lag.
But this indicator is better than mine.

So this is the best oscillator you can find. I am sorry this is not in the public domain.

Here once the bar is calculated it cannot move.,

As for the entry we had a fractal break-out at two levels 15 and 30 m.

The direction is confirmed by two independent algorithms. We may have a singularity like yesterday. Technically the picture is unclear because we are at an important level of support. Waiting for TA clarification I had 40 pips. LOL.

By the way it is possible to argue that you do not need this complex stuff to operate a break-out from a price action perspective. And this is true.
Click to Enlarge

Name: gadalka2.gif
Size: 22 KB

Sixer Nov 24, 2010 4:37am | Post# 854

SSA
 
John,

i can`t find this non-repainting SSA indicator in the elite section of TSD.
Under which name it was published ?

Sixer

michaelwen Nov 24, 2010 7:42pm | Post# 855

The guys from TSD have made a great breakthroughs in the use of SSA.
In fact what I have done previously is to add a digital smoothing of the SSA, so it would not react to a minor change and would react at an important change and in the same time to use 20 bars lag.
But this indicator is better than mine.

So this is the best oscillator you can find. I am sorry this is not in the public domain.

Here once the bar is calculated it cannot move.,

As for the entry we had a fractal break-out at two levels 15 and 30 m.

The direction is confirmed...
John, for me ,I still like your FullSSA-Normalize . yes, it is repaint a little bit. but I am not use as my entry .and I am not use as forecast. I am only use its slop as momentum indicator. only trade at same dirction with it , not against with it`s slop. I check with TSD elite ,from pircture of SSA Squeeze of Averages indicator, looks nice , wish someone can post in here, so we can test it. keep good work!

michaelwen Nov 24, 2010 7:43pm | Post# 856

John,

i can`t find this non-repainting SSA indicator in the elite section of TSD.
Under which name it was published ?

Sixer
I belive call : SSA Squeeze of Averages

wish someone can post in here, so we can test it.

Sixer Nov 25, 2010 1:06am | Post# 857

SSA
 
michaelwen,

thanks for the exact SSA indicator name.

Sixer

John Last Nov 25, 2010 1:50am | Post# 858

SSA
 
This indicator is not in the public domain. It is open source but it does not mean that it is free to post in an open forum. We have to consider the work of the coders.

On the other hand the idea should be open so we all can benefit from a positive black swan, some coder being able to make it better.
To the greater benefit of all.

The basis of the indicator idea of coloring is this code:
http://codebase.mql4.com/3389o

The library of SSA is different. I am not aware if it is different from inside but the files are file size is different. I believe that the free library can be used too.

So the SSA normalize adds arrows on the graph that do not change when the recalculation is done.
So somewhat instead of arrows a better idea is to have bars that will not change once they are closed.

Another hint this can be done even better because you can add the jjma function or any other digital smoother, instead of EMA smoothing and HMA.

When I add a jjma function of the original normalize_SSA I have nice results because it slows down a little bit the movement but moves when it has to. It is not then a pure SSA normalize but somewhat hybrid indicator.

purley Nov 25, 2010 3:26am | Post# 859

John ,

I recently found this indicator AFIRMA(Autoregressive Finite Impulse Response MA)addition to SSA it can be great tool ,red part is recalculating constantly , I wonder if it is possible to stop recalculating constantly with end point or some other techniques. http://codebase.mql4.com/ru/970

http://i51.tinypic.com/zx8sxd.png

yourspace Nov 25, 2010 3:48am | Post# 860

all of the perfect cycles are repainting or recalculating


© Forex Factory