Digital ACSTrend - page 6

 

Changing period

This was a concept and you are able to test it. Please consider all this as concepts and ideas that are not proven (anyway we cannot prove really anything).

The idea is that we are able to adapt the Oscillators to the cycle period. And if we treat a Neural Net as an indicator, why not ?, we could train it too to the cycle period. However a Neural net is able to explore a bigger period for example 1.5 of dominant period, or 2 dominant periods, you can vary manually this.

However keep in mind that the cycle period are lagging in their estimation but it is OK. Of course I use also a third party software of Hilbert transform. However I post this because I wanted something free for you. And the results do not appear really different. We should explore other methods if they are available.

This is a hybrid strategy we use a methodology from Ehlers but we apply a Neural Net.

My idea is based on the following ideas:

-we loose information with every bar because of the chaotic nature of the market.

-even if we do not loose information the neural net are not able to find a global solution they they get "sucked" to the local optima.

So I wanted to use a neural net on a shorter period but I do not wanted an arbitrary period. And logically there comes the combinations with the cycles.

The theory is that as the price series are chaotic we have different kind of attractors out there hidden in a multidimensional phase space. In theory, the neural net is going to look where they are hidden and try to find them. And doing so we do not need to go very far. (Even if we go the results will be pointless because the market is not a mathematical function we can approximate, it is not).For me it is a big misconception that we can "approximate" and "generalize" a solution through a specific design of Neural net architecture.We can't just because the market is not a function! However we still can use the technology but we should know its limitations.

And there is a hypothesis that the fractal attractors produce the cycles we observe and we can measure. All we can possibly do is to approximate the influence of the attractors. The neural net is showing where the price is attracted by the attractor. That is my hypothesis.

Well the signal changes it mind very often, and that is true the market changes its mind very often too. So the best use is when we do not have a lot of volatility.

 

This is a shot of today. Well, we have a confirmed signal by the ASCtrend signal and stops.

However the iVAR is telling us another story we have ranging patterns and high fractal dimension. The principle is that unless there is a fractal break - out we will stay there.

The idea is to not take any directional indication when we have high fractal dimension. We need a low fractal dimension to take directional indications.

You can solve this by going in lower time frames wherein you can find low fractal dimension and use directional indication strategies.

Or you can use a proved range - bound trading system (and this thread does not covers that).

The neural net is telling that the market is going sideways.

The linear regression is showing that we are in the limit if the up channel. Well if you add a second order of polynomial you will have another picture, but I am a fan of the linear regression (for everything else I prefer the SSA).

The BB Caterpillar Alma is showing that we are really in the border line.

Summary:

Unfortunately by now we cannot do any better than the conventional Technical analysis, we are in range mode and we have to wait for a break - out.

Remember the idea: When in high fractal dimension we will have a break - out in lower fractal dimension (combined with technical break - out). The directional indicators do not work by their design (for example you need to wait the close of the bar for ASCTrend and that would be too late).

When in low fractal dimension we need to follow the directional indicators.

Files:
alma_1.gif  33 kb
alma2.gif  32 kb
 

High Phase Space Singularity

This is an example of typical price action when we are in a High Phase Space Singularity.

The phase space of the possible solutions is very complex. It is so complex that the directional algorithms are not able to find a correct solution in our time of reference.

In order to have a trend, one of the new forces on the market is the cooperation of trading algorithms. The Trading algorithms have a different psychology, when they find a solution they cooperate (the humans have a non cooperative behavior).

Here we had none of this because of the complex phase space. Look at the picture the fractal dimension is high in different time frames.

And so the price action in those conditions is ugly, jagged, the price does not know where to go. This is really typical. An experienced trader would notice and say. "Look how ugly is the price action this is a difficult market." The normal trader will draw his trend lines and will make predictions in particularly difficult conditions.

Of course when we see the picture we see that sometimes our algorithms are able to find a solution but this is coincidence and by chance.

And the money management in those conditions should be.

Once a winner always a winner.

The other guiding principle is to use large stops and little leverage.

I tested once to open several hedging orders with no directional indication whatsoever. And then I started to close them one by one once one by one they become profitable. It was fun. Those kind of strategies work really well in those conditions. (The opposite is also true you will know when those strategies may terribly fail when we have a low phase space singularity and the price drives exclusively into just one direction).

I can make the following comparison. Using and taking into account the fractal dimension and its effects is like counting the cards in Black Jack. You know what cards are being played and you can make the most probable bets. When you see a high fractal dimension it would be stupid to bet on a trend.

Now the fractal characteristics of the price are not typical for an impulse trend movement. When we see a fractal break - out we will know what comes next. Well when we see a low frame it looks like a trend but it is not really by now. The price can take this entire distance in minutes (seconds) and go downwards. I consider those moments like the breathing of the market. The market as a living creature breathes and has its own rhythm.

Files:
fractdim.gif  33 kb
fractdim2.gif  36 kb
 
John Last:
Hi, There is a RAR file.

Extract the file into a normal dll file.

This file need to be installed in the experts/library folder.

This is a mql base indicator

Next price predictor using Neural Network - MQL4 Code Base

Please read the thread there. This is a working neural net implementation for Metatrader.

However it keeps crashing all the time. So a third party fix was prepared and I have posted the fixed version.

BPNN dll temp fix - MQL4 Code Base

My contribution is to constantly adapt it to the cycle period. I am still testing it. Sometimes it gives really interesting things.

The next idea is to use the "output" into other indicators.

In fact you will have an indicator which will show you what is predicted to the close of the opened bar.

Until now many uses: FullSSA with this looks promissing (I call it FullNSSA), however I am still testing it.

I've put these indicators from post no #40 -

BPNN Predictor Cycle Period.mq4 (9.5 KB, 29 views)

File Type: mq4 CyclePeriod.mq4 (3.9 KB, 22 views)

File Type: mq4 BPNN Predictor.mq4 (9.5 KB, 26 views)

to the indicators folder and unpack the file

BPNN copy to library.rar (6.5 KB, 28 views)

to the libraries folder...

and ... i see no results. What is wrong?

 

temp fix version BPNN is not work.

I did try to use temp fix version of BPNN. It is not work, come out "error 126". I believe the parameters pass does not match. Need to be modified???

The original one is workable, but, get crashed on demo account often. Sometimes, can see red line on chart, if don't close the crash message windon, it may still keeping running there. So, it is a problem when return to main subroutine. Some objects in c++ may be not looked after well, or use wrong objects, or message passing to objects may be wrong, or object fail to react to the message, or object inside may be wrong.

 

Crash

I am sorry to hear that. Well did you try with the original BPNN? Does it work?

Another thing maybe if you change the platform and use another MT4 with another broker it may work. For example I use FxOpen. Sometimes the fixes are not so obvious.

 

Fractal Break - out once again

Here I want to show the concept in practice.

Yesterday we discussed the High Phase Space Singularity, when the market really does not know which direction to take.

Yesterday I wrote that really we do not know what is going to happen after High Phase Space Singularity when we try to make a precise local prediction.

We needed to wait for a fractal break - out. And the fractal break - out occurred. It had to happen. And the market went higher. Now when you look at the graphic you may think well, we had an established trend from the break - out of the trend line of the down trend. Well, well afterwards it is always easy to analyse.

I mean so because many say all that correction as a sort of correcting pattern, before a new downtrend impulse.

However as I telly you the High Phase Space Singularity is a shitty place to make predictions especially for the future.

The Fractal break - out gave me an opportunity for a long position, and this position was highly counter - intuitive for me yesterday, I was internally struggling to take that trade (I am very shitty Technical Analyst with huge baggage of bad luck).

OK I propose to do the following. Open your history and drag and drop your loosing trades. Do they happen, many of them when we have High Phase Space Singularity.

I add a shot of the normal ASCTrend and you will tell me in what fractal geometry context are the best signals (I hope you can see). Exactly for that reason the optimizing does not work as good as we may think. Because what really matters is the underlying structure of the fractal geometry of the price. Unfortunately it is not that simple. On the left part we had bad signals many of them on the right part just several but really nice signals the settings are the same.

Whatever indicator, again whatever indicator, you may use it is important to know if it will work by its design in the fractal geometry context you take the signal.

Files:
dalgi.gif  25 kb
idea.gif  23 kb
 

JStochastic on MESA

Here is a shot of this. I do not post it because it does not work properly yet.

You can compare with the adaptation according to the cycle period indicator.

Files:
idea2.gif  25 kb
 

Hey John

What settings you put and what indicator you use (ama or alma ?) on the screen of eurusd 1 hour, 1 post before ?

Thanks !

 

Settings

Remember the Simple Moving Average is hard to beat LOL.

This was the normal ASCTrend with its standard settings of Risk 3 and Money Risk of 1.5.

And on this shot is my version with Mama.

With the same settings but in addition we have

Fast limit 0.2

Slow Limit 0.05

Once again you see that when the FGDI is below the 1.5 line and is red (iVAR below 0.5) we have periods when we have nice signals (well we do not know if we will have a trend but during this time the directional algorithms perform at their best).

The best time to trade is when the FGFI makes a fractal break - out and is pointing down. When it is extremely low very often we have a Black noise reversal as it is predicted by the neural net that we can expect something like this.

Files:
dalgi2.gif  32 kb