Optimized Trend Trading

I would like to answer your question but main idea was to share the fractal break - out ideas.

I have used Neuroshell Day Trader with the metatrader datafeed. What I like with it. I like that I can implement really quickly trading ideas.

What I do not like. I do not like that it is a black box, you have to pay for, well some people do not but this is another story . Of course I know that it uses some kind of proprietary cascade correlation algorithm. They call that turboprop, and it is maybe a cousin of the quick-prop algorithm.

And the neural net have limitations (of course).

If you are interested in investing in it, I can recommend the Products of Noxa analytics as the CSSA (this product was commented a lot here and other treads).

I want to emphasize here its relation with the fractal dimension. A member has observed that the CSSA give a lot of good signals in particularly noisy environments but when a sharp shift in the market occurs the CSSA loose the fit.

So there come the idea to constantly monitor the market with the fractal index and I know that such a shift occurs when there are a strong and violent variation of the Hurst exponent. So you use your net but you have a panic button.

My observation of the Neuroshell and the other nets is that they really cannot predict break - outs. So my opinion the best use of the neural nets are in relatively calm market environments when a break - out is unlikely to occur. You learn very well your retail provider algorithm LOL.
I think a lot of neural net practitioners have faced how a neural net counters the trend .

In fact you may laugh at my reductionist approach but a neural net in my opinion is just a better and nonlinear indicator. It looks like an advanced non linear oscillator (in fact the Neuroshell guys do not try to predict price levels, they predict percentage change in open).

I think the best that can be done on a retail level is to combine SSA with Support Vector Machines and use it in a calm market environment. There are several open source softwares that permit that combination. That will be my range bound market system.

If the price series are clearly persistent with medium to strong volatility I would prefer a digital filter optimized with genetic algorithm.

I want to add something more concrete. This is a trend following system. Basically this is the volatility system of Wilder but with digital filter smoothing. My modest contribution is to add the phase as a parameter too.

OK just to remind install the mqh (jjma series) file in the include folder.

The default jma is 1. Put it to 6 and you will see the difference. Play with the phase too. Good luck.

NOT YET CAMARADEN, NOT YET

I will add three comparative shots.

Smoothing of 6 with three types of phase:
-100
5
100

Do you see the difference? ENDE

Addition to being a trend indicator Mark Jurik suggesting to use fractal dimesion(or cfb) as the speed adaptor for momentum indicators.

OK, I will show an example from today. Well folks that was not my best trade but this illustrates the setup and you can do better.

So we have a fractal dimension higher than 1,5. Then we have a change, the fractal dimension is getting lower than 1,5. Visually he FGDI goes from blue to red zone.

In the same time we have a positive Hurst difference, or a Peak in the Hurst difference.

So this is the setup.

On the shot you can see the setup is confirmed by my trading system as a safe entry.
The SSA and the PFE are showing negative momentum.
And the trend strength indicator is going up.

In fact the price series follow a cycle between a fractal dimension below 1,5 and above 1,5. And those cycles are antipersitent as the volatility is.

The very good point of this setup is that when you see a break - out there is almost always a buffer time and you can use this time to enter at normal market conditions, without slippage, re-quotes etc.

If you miss the setup and the normal market conditions look for a correlation in the pairs. For example often when the Euro goes up the USD/CHF will go down (negative correlation). There are indicators measuring the correlation, but take care if the correlation is not working check the fundamentals what is going on.

i put the mqh file in the include folder and all other indicators in the experts/indicators folder and could not get it to work. Am I installing it improperly?

Copy the files

experts/include - here put the mqh file
experts/indicators - here the mt4 files
restart the metatrader you should find the indicators in the custom indicators. Modify the smoothing to six for example.

I use Metatrader 4 with 5 digits broker.
(NOTA BENE When you start the indicators allow dll and extern experts otherwise it will not work, this is obvious by many forget this detail)
I checked on another PC and I have found no problem m8.

This is a very sound trend following system. Do not try for scalping unless you know exaclty what you are doing.
The best way to go is to program an expert and with the genetic optimizer to find the best solutions for the actual market state.

And that is true even you are doing manual trading because the genetic optimizer will tell you what are the best set of solutions: smoothing and phase.


The best market conditions are:
FGDI: 15 m : below 1,5
FGDI: 30 m: below 1,5
That means that we are in persistent market conditions. Do not predect, follow the digital filters. And those are good.
The algorythm is capable to find a solution in very difficult conditions, but do not count on it very often.

I use some additional indicators to filter the market conditions.
JCFBaux - as a trend filter: make google search
I also use a normalyzed version of a SSA (singular spectrum analysis) and a modified version of polaryzed fractal efficience.

who let the spammer out,who,who who

with all these indis how much price action can u actually see

This is though question.
This setup is used to confirm a break-up. Every Break-up is characterized by a change of the fractal dimension characteristics.

I have not observed a false positive so far (i.e. having a real break - out without change of the fractal dimension).

On the other hand you need to make a statistical study how much of the fractal break - outs come to a real break - out.

A friend of mine did an expert, a conceptual one: who is taking action when the fractal dimension drops from 1,48 to 1,46 taking the position in the direction of the polarized fractal efficiency. I did a simple R/R rule to 1 to 1 of 15 pips with time filter from the open of the European session + 4 h, after that all positions are closed no matter the result.

The setup was profitable only on the EUR/USD. And it makes money, but really not much. The fun point is when I made a study for the last 10 years it was profitable but imagine the winners were 51 - 52 % (This is giving us a slight edge). This year the ratio of profitable deals increased to 60 %.
In fact the Hurst exponent and the fractal dimension is giving a slight edge.
When you have a transition between 1,48 to 1,46 of fractal dimension you have maximum 54% percentage of chance that the next movement will be in the same direction (a simplistic view of the Hurst exponent is to watch it as a measure of the probability that the following movement will go in the same direction, for example if we have Hurst = 0.5 (it is 50% probability to go in either directions, if Hurst is 0.6 we have a 60% to go in the same direction (persistent movement), i we have a Hurst exponent = 0,45, the probability is 45 % to go in the same direction and 55 % to reverse (anti-persistent caracteristics)

Remember: FGDI and the Hurst exponent are connected :
FGDI= 2 - Exponent of Hurs.

One way to increase your chances is to use this setup in period of high expected volatility. There are online ressources with that information: basically the open of the European and London session.

So if we have a winner the winner is a big one. In the shot is an example of big winner. I used a fixed objective of 30 pips I think.

How to deal with that setup?
I think that a setup is a setup after your position is opened you need to make a prediction and adapt your money management according to that prediction.

In fact this setup is not intended to replace the technical analysis but to give
a new level of objective perception.

As a prediction I have a hypothesis. After the fractal break - out, the price makes a first correction, the price fails to do for the first time a higher high (in the price action terminology this is a Ross Hook).

The idea is that this is revealing information how much the price will correct. This is classic the Dinapolli objectives of fibo expansion. But the point is to measure the low from the point of the fractal break - out and not from the point of the low. Why? This is bacause I want to measure something between two events. The low can be an event but not necessarily, the fractal break - out is the true event.

The theory is:
After a fractal break - out a new information is coming into the market
The market is out if balance and is perturbed. In that moment the neural nets tend to fail. So the fractal break out sets the initial conditions. After that perturbation the market seeks a balance and tries to correct. It makes the first correction. A measure of the fibo expansion is a simple tool to set a point attractor to the price (the simplest form of an attractor) (I am sure that there are others more scientific tools). And if the market breaks above after the first correction and the fractal dimension is still going lower we should use our objectives. If the fractal dimension reverses and goes higher the chances are that the market will not make a sustainable trend.

After the initial break - out the market generally makes new fractal shifts to downside ( I call them a break - in, the FGDI goes inside its territory. e.g. being low, he gets lower) (each of them is adding new fuel to the movement).

A good idea is to gave a good algorithm who will guide you in this process.

The code posted here is very sound, I believe at the limits of Metatrader.
(NOTA BENE When you start the indicators allow dll and extern experts otherwise it will not work, this is obvious by many forget this detail)

When we have a persistent movement every strong "kick"in the price can provoke it to change its direction. It is referred as a black noise phenomenon. This phenomenon explains a lot, this provokes the V tops and bottoms (and vice versa if you have a high fractal dimension it is unlikely to have a V top or bottom, the time price series usually form a ranging pattern). The price just reacts like a tennis ball.

In this case an algorithm is giving us an edge because it can measure objectively which price hits are strong enough to provoke a hit, able to change the price direction (In fact all the price action guys and gals do that manually and it is possible of course, but difficult because it changes with the cycles). A good approach is to have a digital filter tuned by a genetic algorithm.


I am sorry I have no shots. I hope you find this interesting.
The fractal dimension is giving us a new level of analysis. There are cycles in the fractal dimension that are different from the cycles in the volatility.
We have different fractal dimension characteristics at different levels: 1 m. 5 m. 15 m. 30 m. 1 h. etc. At every level the probabilities are different. The combinations are so many that every day is unique. Everyday brings a new set of combinations. Nevertheless patterns are observed that occur more often that they should do. Those patterns gives us some guidance what is happening and a probability of the future events.

For example a mentioned a point attractor. But there are cyclic attractors too. According to me all the chart pattern are cyclic attractors. All the Technical analysis is just a classification of cyclic attractors without having a clue what really is going on from the inside. And the technical analysis is trying to make use of those cyclic attractors or to make an entry when a shift from one structure occurs. That is why sometimes it works so well and sometimes fail miserably.

That is why the techniques from the DSP as MESA, Band Pass filters, Hilbert transform sometimes work so well and sometimes fail miserably. The cycles are not periodic but they are chaotic.

can you please post your template? I'm not able to make the indicators work

Is there somebody able to make them work?
I am just wondering why it does not work with you.
Anyway I can send the last version without the phase change.
Start them one by one.

There is no change in the files JJMA Series, from the previous post add this file also

The test is if you are able to run the JJMA indicator that means that the include file is installed and working.
If you are not able to run the JJMA there is the problem.

Just one thing have you checked in the Tools/ Menu Options / Menu Expert advisors if everything is checked, Especially external experts

Wow somebody claims that Ehelers strategies are working ??

I evaluated Ehelrs indicators as information extractors in posts 719-724
and it was quite clear that at least in case of such evaluation e.g.
as valuable/not valuable input for NN prediction they were not more valuable
than normal price indicators.

Additionlly I evaluated most if not all of Ehlers published strategies so:

BandPass
Channel Cycle
EhlerChannelPDF
FisherPrice
FisherFriceRSI
HP Fisher
NonLiner
FisherPriceRSI_T
SinCorr

I used Multichars for it and made Walk Forward Analysis. See attachement.
From alll those results you can see than those strategies are not profitable in long term

So my question is: Is anybody able to show min 100 trades OUT OF SAMPLE
and system still profitable using Ehlers strategies ??

See attachement with results

Krzysztof

Hi Krzysztof,

I have been reading your posts with great interest. And here I use the opportunity to thank you for your contribution.

My idea is that Ehlers filter are just good information extractors.
If you have a winning Neural Net model: it will be profitable with Ehlers filters or with just normalized price.

However in particular market conditions they offer a slight edge. My hypothesis is during range conditions they will offer a slight edge.
As trend following at high frames systems they offer a slight edge too. If a normal system works, the Ehlers will perform slightly better. But if a normal system fails the Ehlers system will fail too.

The MESA unfortunately cannot really offer an edge because of the chaotic structure of the cycles. But if you use the MESA in a ranging environment it will help greatly in your timing.

The real problem is to distinguish between the market conditions and use an adapted strategy.

I remember your posts about the failure of the CSSA to work, when a sharp shift occurs. OK, nice but you can observe manually the market conditions.
Just a nice video clip:

http://www.youtube.com/watch?v=723-e...eature=related

In fact I think the best so far I found for range is the SSA + NN. I did not try yet with SVM. Unfortunately I have not the CSSA, so I use a russian version as input behind series of Recurrent nets and Predictive.
The best market conditions are when it is unlikely a break - out to occur: e.g. High fractal dimension and low - medium volatility.
They say never use SSA it repaints. I say I will, I will.

Everything is a matter of perspective, contradictory points of view can be equally true.

not oerking, checked all the ticks on the ea tab

jjma indicator do not load, other do but shows nothing

.

Yes that's true. We can have range conditions which just look like a range conditions but they are just piece of random brownian motion with H=0.5
and in this case all strategies will fail

I think on Jurik site is a article how easy is to fool cycle based indicators
with random serie. They will find cycle anyway

I dont know how you use SSA by definition it makes future leak (repaints)

SVM predictor behave similar to CSSA in case of sharp change of signal but at least don't make any 'oscillations' like CSSA. See this post 58 and down

http://www.trade2win.com/boards/meta...achines-8.html

BTW i evaluated CSSA quite extensively on TRADE2WIN 'Noxa indicators' thread

Krzysztof

Hi,

I made an extensive test for Hurst exponent. And clearly the fractal Brownian motion is an exception and not a rule. And the EUR/USD has the highest Hurst exponent. Moreover I calculated the Lyapunov exponent and it shoved that even from time to time the time price series are not even chaotic. I observed a flat line of Lyapunov exponent.

I saw with my eyes when an experienced trader turned 80 USD into 2400 USD for 8 hours. This was so crazy that his broker published his statement to show that he provides a good service.

Have a look at this article for rainfall predictions:

http://www.iwaponline.com/jh/003/0141/0030141.pdf

And those time series were characterized with guess what Lyapunov exponent .

Really I believe the edge is to know when to trade and what strategy to use when and how. And how to cluster the market states.

And the Peak of the Hurst difference indicator is done by Jean-Philippe, a very nice guy. When you see a peak, you have to be aware that stress tests are coming that may not be good for your system. Really when and how to trade neural nets models is an art by itself, I just wanted more objective approach.

About installation:
Really there is a problem with the JJMAseries file. I feel sorry for that. Make a google search "jjma series" you can find another download location.
Make sure that you allow those files to work. I think this is the problem.

You need to find a way to run the jjma indicator, if you find a way the others will follow.

Download indicators mq4
Copy indicators mq4 to your Metatrader Directory / experts / indicators /
Copy file JJMAseries.mqh to your Metatrader Directory / experts / include /
Start or restart your Metatrader Client
Select Chart and Timeframe where you want to test your indicator
Search "Custom Indicators" in your Navigator mostly left in your Metatrader Client
Left double click on the indicator
Attach to a chart
Modify settings or press ok
Indicator is available on your Chart
How to remove JCFBaux mq4 from your Metatrader Chart?


Here I post a shotbetween the system that is a great reverse engineering achievement asctrend. But here I replaced its heart with gauss filter. See your self the results. It is more reactive sometimes is good sometimes it is bad. But the genetic optimizer will be allowed to explore more options.

So are you able to post result of 100 OOS trades for this system ??

Krzysztof

http://forexforum.bg/viewtopic.php?f=17&t=750

Have a look here. Use google translate.
You are not required to register. You can download the statement on the first page. The account was multiplied 83 times in 8 hours.
25 $ to 2300 $ in 8 hours

Is it possible if the prices were random all the time?

I am sorry the trades were 73 not 100, but this was on a real account.

This is what a good human trader can do if he trades in a good moment.

I do not want to alter the spirit of this tread. The idea is to show that an optimized trend trading requires an appropriate moment for trading.
This trader could never do that if the moment was not appropriate.

Hello;

My mother always says me, you don´t have to work hard, you have to work intelligent

I attach a picture with a simple indicator based on (Close[i] <> Open[i])

Similar results like DSP.

With all my respects, at this moment NO RESULTS.