Digital ACSTrend - page 31
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Future developments
I see this thread as a complex of some independent projects:
1. Using digital filtering for enhancement of ASCTrend sig.
And that is the digital ASCTrend. Some of the algorithms remains in the original ASCTrend.
2. Using digital filters from Jurik type digital filtering by Kositsin with cycle period adaptation according to the DSP theory of Ehlers.
I used the free library of Kositsin.
-JMA_WPR with Cycle period adaptation
-JStochastic with Cycle period adaptation
-JMA_WPR with MESA adaptation by Richcap Mesa
-JStochastic with MESA adaptation by Richcap Mesa
3. Use of different additional filters mainly:
SSA based
I would especially than Mladen for the fast end - point version of the SSA, the SSA end - pointed
There are different mods (with histogram without etc.)
I used an adaptation by the fractal dimension, the volume and entropy.
Digitally smoothed Entropy based
This is the original work of John Conover.
The digital smoothing by jurik type of smoothing (according to igorad it is a triple smoothing: EMA, Kalman filtering, and additional smoothing using volatility) offers some very interesting enhancement
Mesa adaptive digital filters by Richcap
Lately ,the rise of the more powerful PC allowed a more frequent adaptation of the original filters making a boost in their performance. I just pointed those indicators because they deserve it. Mr. Richcap made some major contribution on the digital filter methods.
There are some other filters.
I can add the Lyapunov which looks like some enhancement over the traditional momentum.
4. Trend classification algorythm
Here the idea is to replace the original trend classification of ASCTrend (the thing with the dots) by Brain trend SSA ep. Here and on the original thread some digital ASCTrend were prepared. The original ASCtrend is based on SMA, which are really hard to beat. Anyway the Mr. Tools MAMA mod was probably the best digital analogue.
5. The market state paradigma and Chaos analysis
I think this is very important. One of my basic goals was to gather the available information and to post it here.
Look at the previous post for more details.
6. The predictive edge
I added some tentatives for a predictive edge. But this is really not serious.
I mean the cycle adaptation of the BPNN or the Radial basis mods.
What I meand for future development is to use Rapidminer and exporta data by an expert to Rapidminer.
As an introduction look at the tutorial videos for Rapidminer 5 (Rapidminer rocks because it is free):
http://www.neuralmarkettrends.com/tutorials/
/you do not need to go there but it is necessary if you want to take the maximum from the tools here, you need a week - end to play with. I can post the ready code but the idea is to learn it and build your own model./
The idea is that we have some very nice algorithms here for data pre-processing. I want to use advanced data - mining algorithms to use them at their maximum performance. The Support Vector Machines are my choice and I am their fan.
The result would be analogue to the results you may achieve using expensive commercial softwares (it may even be better as they do not implement more advanced algorithms as Support Vector Machines ). Well, even if RapidMiner is free the knowledge and the courses how to use it cost more that the commercial software LOL.
The relation between MT4 and Rapidminer could be very interesting. That is because by now under MT4 we have a huge collection of very powerful algorithms. They can be used at their limits only if they are exported into state of the art data mining software.
This would close the thread from my part. I wanted to make something more organised and structure /I even consider to add bibliography as in a serious scientific work/.
Thanks
Thank you Richcap for the nice comments. Without your MESA implementation some very advanced mods are not possible.
I understand this forum as a gigantic mega mind. Any input is important even the most unpleasant. This forum is a very powerful research network.
When you release a set of indicators the input by other guys is valuable. The problem is that it may come even after years.
I am believer in the Positive Black Swan Phenomenon, that sometimes great systems just happen to be created by chance, or are the product of a sudden intuition. I personnaly experienced that when I saw the Mr. Tools mods, who had the idea to use double smoothed jurik of Mladen. Or Camaron who found the best default settings for the Brain Trend SSA ep. Or Jaguar which work on the Lyapunov made some things possible.
Or last but not least Mladen who decided to release to the public the End - pointed SSA with fast processing dll. All that happened by chance, or the process was chaotic
.
Some little mod or hint may change a lot of things.
Maybe a combination between releasing and keeping private is the best.
When you release you may benefit from the black swan. But sometimes you just do not want to release or you use some other peoples work so that is not possible.
In fact I believe that the forum is just the surface beneath there are a lot of private research groups who collaborate on their personal projects and keep things private. That is completely normal.
That is my answer when my girlfriend asks me if I am crazy, releasing systems for the others and putting too much effort on working on them.
In my case I have also some unreleased systems. But curiously the released systems just made the best performance.
Here I expressed some of the most crazy hypothesis, other things may be called just fringe technical analysis.
I just want to use the instruments we have to their very limits.
table of correlation features
Hi all
While investigating on correlation matrix, I found a pdf file, quite interesting
If I understand well this document, It seemed input datas need first to be purified by using a table of correlation and eliminate wrong values
I put this file in attachment for your knowledge
Why ? I hope the attributes will be used more efficently inside RapideMiner i.e
One of these days you will see something very weird in this forum.... lol
check the .rar file
Rgds
the largest Lyapunov exponent
Well well
After working hardly and harshly and created some tries (more or less sucessfull), and as I guessed before, it should be better to fetch the largest Lyapunov exponent directly from the artifical neural networks than from a specific metatrader indicator.
I found this, talking about this subject.
Anyone got an idea ??
Rgds
john
Addionnal note
This kind of problems is not due to SSA end - pointed, which is perfectly casual indicator.
It is due to the carrier. Any improvement is welcome, because it is beyond my competence and skills.
Warning
Unfortunately the bt SSA ep is unreliable. It breaks down very often. In order to know what is going on you need to run the tester with a blank expert, or one of the default experts and run a visual test mode.
In that way you can see a visual representation what is going real time.
Even if the SSA ep is a casual indicator, a repainting may occur (I mean in the Brain trend SSA ep). The calculations are heavy and during high volatility it may break - down.
Even if you use different brokers you may have different results.
Every help to make those instruments more reliable is welcome.
Here I add two screen shots the next day on 15 and 30 m chart. On the 15 chart the signal from yesterday is displaced several bars earlier.Even if it was correct you should take into account this. I do not know why that happens, it is the code. On the 30 m chart it looks to me the same.
And if you change the number of the bars to look the results will vary too.
A friend of mine did an expert which on the back tests showed 2000 pips for the last year. However when I plot on the chart the actual signals are inconsistent with the indicator. I just cannot make it work.
On the 30 m chart it is the same with no difference.
I see this thread as a complex of some independent projects:
1. Using digital filtering for enhancement of ASCTrend sig.
And that is the digital ASCTrend. Some of the algorithms remains in the original ASCTrend.
2. Using digital filters from Jurik type digital filtering by Kositsin with cycle period adaptation according to the DSP theory of Ehlers.
I used the free library of Kositsin.
-JMA_WPR with Cycle period adaptation
-JStochastic with Cycle period adaptation
-JMA_WPR with MESA adaptation by Richcap Mesa
-JStochastic with MESA adaptation by Richcap Mesa
3. Use of different additional filters mainly:
SSA based
I would especially than Mladen for the fast end - point version of the SSA, the SSA end - pointed
There are different mods (with histogram without etc.)
I used an adaptation by the fractal dimension, the volume and entropy.
Digitally smoothed Entropy based
This is the original work of John Conover.
The digital smoothing by jurik type of smoothing (according to igorad it is a triple smoothing: EMA, Kalman filtering, and additional smoothing using volatility) offers some very interesting enhancement
Mesa adaptive digital filters by Richcap
Lately ,the rise of the more powerful PC allowed a more frequent adaptation of the original filters making a boost in their performance. I just pointed those indicators because they deserve it. Mr. Richcap made some major contribution on the digital filter methods.
There are some other filters.
I can add the Lyapunov which looks like some enhancement over the traditional momentum.
4. Trend classification algorythm
Here the idea is to replace the original trend classification of ASCTrend (the thing with the dots) by Brain trend SSA ep. Here and on the original thread some digital ASCTrend were prepared. The original ASCtrend is based on SMA, which are really hard to beat. Anyway the Mr. Tools MAMA mod was probably the best digital analogue.
5. The market state paradigma and Chaos analysis
I think this is very important. One of my basic goals was to gather the available information and to post it here.
Look at the previous post for more details.
6. The predictive edge
I added some tentatives for a predictive edge. But this is really not serious.
I mean the cycle adaptation of the BPNN or the Radial basis mods.
What I meand for future development is to use Rapidminer and exporta data by an expert to Rapidminer.
As an introduction look at the tutorial videos for Rapidminer 5 (Rapidminer rocks because it is free):
http://www.neuralmarkettrends.com/tutorials/
/you do not need to go there but it is necessary if you want to take the maximum from the tools here, you need a week - end to play with. I can post the ready code but the idea is to learn it and build your own model./
The idea is that we have some very nice algorithms here for data pre-processing. I want to use advanced data - mining algorithms to use them at their maximum performance. The Support Vector Machines are my choice and I am their fan.
The result would be analogue to the results you may achieve using expensive commercial softwares (it may even be better as they do not implement more advanced algorithms as Support Vector Machines ). Well, even if RapidMiner is free the knowledge and the courses how to use it cost more that the commercial software LOL.
The relation between MT4 and Rapidminer could be very interesting. That is because by now under MT4 we have a huge collection of very powerful algorithms. They can be used at their limits only if they are exported into state of the art data mining software.
This would close the thread from my part. I wanted to make something more organised and structure /I even consider to add bibliography as in a serious scientific work/.J-P Poton, Fractals, Technical Analysis and other things..., has a fairly fresh entry in his blog, "The possibility of cognition", and I believe what he writes in paragraphs II and III are very true in terms of what we have been looking for/ought to look for, being the "Fractal aficionado" that I am. What do you think guys?
Some experiments
Very interesting post of Mr. Poton indeed.
I just want to add some experiments with the SSA bands. Withle the blue lines we have a repainting SSA bands. It is beautiful and it is the Hurst dream.
However the question is what we can get from casual bands. Can we get the same patterns using casual instruments?
My answer is that it may be possible but we cannot have the beauty of the repainting SSA.
Please consider some shots
The caterpillar bands were posted on forex factory and they depend on Caterpillar indicator.
Those instruments repaint but they could be used on offline charts as analytical tools. Use the caterpillar dll.
Forex Factory - View Single Post - Optimized Trend Trading
Check this article and you would know some ideas.
Forex Factory - View Single Post - Optimized Trend Trading
indicators
I posted the link on the FF because I did not want to flood this thread with repainting indicators (still of some use of course but it is expert's domain not for the common user anyway).
I just want to point that the idea to use the Mladen's end - point was not exactly mine, it was suggested by a forum member. He can reveal himself.
As I am incompetent in coding I used the Mladen's code and I modified just one line at the end.
and
So there are two separate indicators, one is giving you the upper channel and one is giving you the down channel. Well it can be done easier with all into one indicator.
The center line is just the SSA end - pointed in this thread, and of course you need its library.
https://www.mql5.com/en/forum/180729/page9
I posted the link on the FF because I did not want to flood this thread with repainting indicators (still of some use of course but it is expert's domain not for the common user anyway).
I just want to point that the idea to use the Mladen's end - point was not exactly mine, it was suggested by a forum member. He can reveal himself.
As I am incompetent in coding I used the Mladen's code and I modified just one line at the end.
and
So there are two separate indicators, one is giving you the upper channel and one is giving you the down channel. Well it can be done easier with all into one indicator.
The center line is just the SSA end - pointed in this thread, and of course you need its library.
https://www.mql5.com/en/forum/180729/page9Hi John,
I put two into one for you