Digital ACSTrend - page 30

 

Preferred System Now

John,

Thank you for starting this thread. Very interesting.

Question for you: what is your preferred ASCTrend system at the moment based on your trading experience/backtesting?

JSchoastic based?

Could you summarise all the components and explain which parts are still work in pogress and which are only available in the Elite section?

I'm looking to backtest / develop the system, so it would help to know what to work on

kind regards

f451

 

Preferred System Now

Hi Again John,

I'd like to second F451's request as I have followed this thread for some time and want to begin testing and evaluating the system. Your comments have been very beneficial.

 

Thanks

John Last:
Hi,

I just do not have the time. You should start from the beginning and see the development of the ideas. The ideas and the indicators go together.

So this is not really mechanical. I have chosen not to give you rigid rules or settings, because that is a responsibility I cannot take. Those are instruments and use them as instruments.

That makes this thread very different from some others where precise rules are given. (for example take the trade on this frame based on the indications of the indicators x, y, z on the other frame etc.) There is none of the kind.

However basically there are three ideas:

1. A classification algorithm

Here there are many of them, I use mainly bt stop ssa ep.

It answers the question. The trend is up or down

(the bars answers the same question the green bars show sideway market, but we do not have adaptive bars yet.)

2. An entry algorithm

ASCTrend CyclePeriod, ASCTrend JStochastic and ASCTrend MESA, or even the normal ASCTrend sig.

It answers the question. Is it a good place to enter.

3. Market state evaluation method

-fractal dimension (and the subsets with some hypothesis)

-entropy

-Lyapunov exponent

-Kurt Faith method combined with fractal dimension

We miss the correlation dimension

and all that does not replace the classic: volatility estimation, and volatility based on the time of the day.

e.g. chart patterns etc. etc. etc.

Just look how our classification algorithm worked on daily charts.

That is why I posted this beast here to ask you is it true?! Is it too good to be true?!

Anyway do not bet the house on this LOL. Always, Always, Always use good money management and risk control procedures.

Bear in mind that I did not optimized that, I just plotted the default settings.

Thanks John - I'm not after mechanical trading rules - I'm currently re-reading the thread again to make sure I've covered off the issues and the current options for each component in this development of the ASC trend system.

For the parts that are implied, and not covered in the thread, please correct me if I'm wrong:

Volatility estimation - ATR is a reasonable estimate this forms the basis of the position sizing and money management

Correlation - knowing the corerlation among the forex pairs, so your trading risk rules can be applied. Entering multiple positions across instruments (e.g. long USD pairs) could be inadvertantly breaking risk control depending on the instrument correlation.

Market State evaluation - choice of Fractal Dimension or entropy to determine if stable or volatile and trending or ranging. As I understand it we want to trade Trend ASC in trending markets and get conflicting signals from ASC Trend Entry Signal generators (section 2 above) in ranging markets.

One question around the classification algorithm - what do you mean by adapative bars?

regards

f451

 

Hi,

I just do not have the time. You should start from the beginning and see the development of the ideas. The ideas and the indicators go together.

So this is not really mechanical. I have chosen not to give you rigid rules or settings, because that is a responsibility I cannot take. Those are instruments and use them as instruments.

That makes this thread very different from some others where precise rules are given. (for example take the trade on this frame based on the indications of the indicators x, y, z on the other frame etc.) There is none of the kind.

However basically there are three ideas:

1. A classification algorithm

Here there are many of them, I use mainly bt stop ssa ep.

It answers the question. The trend is up or down

(the bars answers the same question the green bars show sideway market, but we do not have adaptive bars yet.)

2. An entry algorithm

ASCTrend CyclePeriod, ASCTrend JStochastic and ASCTrend MESA, or even the normal ASCTrend sig.

It answers the question. Is it a good place to enter?

3. Market state evaluation method

-fractal dimension (and the subsets with some hypothesis)

-entropy

-Lyapunov exponent

-Kurt Faith method combined with fractal dimension

We miss the correlation dimension

and all that does not replace the classic: volatility estimation, and volatility based on the time of the day.

e.g. chart patterns etc. etc. etc.

Just look how our classification algorithm worked on daily charts.

That is why I posted this beast here to ask you is it true?! Is it too good to be true?!

Anyway do not bet the house on this LOL. Always, Always, Always use good money management and risk control procedures.

Bear in mind that I did not optimized that, I just plotted the default settings.

Files:
signal.gif  32 kb
signal_gbp.gif  28 kb
 

Processing information

Well I will take care not to answer being pressed by the time.

But you say very important things especially those unrelated with the discussion so far about the correlation.

1. I was always surprised when some analysts recommend long on EUR/USD and short on GBP/USD or vice - versa, when clearly they are in almost perfect correlation at that moment.

People need to get aware of correlation however a reasonable thread cannot cover all the important issues. We can just mention and give appropriate links of interest. My studies (unconfirmed hypothesis) showed that the evolution in the time between EUR/USD and GBP/USD shows two phases. The first phase is a deterministic evolution, the second phase is chaotic but more chaotic than the time series by them selves. So there is a constant switch between these two regimes.

2. Another topic is how I process information:

I use the following:

2.1 market condition analysis

-fractal dimension

-entropy

-Lyapunov exponent

Now with the link mentioned of the Google books we have a complete chaos theory analysis methodology. We miss only the correlation dimension plot.

-volatility (predicted volatility, news, time of the day)

-existing correlations

-market statistics and the position of the VWAP

-information on market orders and their levels

-options barriers

-fundamental sentiment : I am not a specialist so that means that I read information but I do not make fundamental analysis by myself. A modern approach is to make a data meaning of the fundamental information.

2.2 Check of the existing model (this is human pattern recognition)

-technical analysis pattern recognition (the automated software analysis is interesting there is a pattern recognition scanner, I think that Saxo bank include this kind of analysis on their platform)

-fractal pattern recognition (fractal break - out)

-fuzzy market state recognition according to the methodology of Kurt Faith

-other things (it may sound curious but I have build a model using Fibonacci and Gann lines that worked very well for the last few weeks, however all that is off topic and I do not discuss this stuff here) (also I like to identify Elliot waves models but in fuzzy way, e.g. the impulse has a low fractal dimension (and the simple A-B-C) the more complex corrections have a high fractal dimension) so even if it is fuzzy it is objective, by the ways I prefer the Glenn Neely method, but I use a simplified approach.

3.3. Whatever model I do I should not enter into position unless I am allowed to do so. Well why models? The models give me some kind of psychological feeling that all the time spent on this is not in vain and bull shit and I am more confident for bigger targets.

Trend Classification algorithm: Brain Trend SSA ep

Predictive trend classification algorithm: Support Vector Machine model

Cycle algorithm: I am focused on micro cycles: ASCTrend mods

Predictive algorithm: some Neural nets models

/as for the bars I means ASCTrend1 bars but to include adaptability if possible, as I am not a fan of colored bars I missed that so far/

This explains where I am and it is an ongoing amateur research. The high - frequency professionals measure by now the delay in nanoseconds.

 

A Good Starting Place

Here, makes sense, or on a new thread if preferred,

would one of the members who has closely

followed this thread and compared the tools

shared so far, as well as taking the suggestions

from the thread starters last post here before mine...

would you post a sample setup compiling these

tools into a setup?

An H1/ H4/ M15/ Daily.. most good systems as such

can start from the same base setup.

It would be greatly helpful I believe in getting

this process to the next level, ie real demo

and live trading to fine tune the "method"

All the best

Brian

 

Ideas

Basically look at the daily. This is my favorite time frame.

BrainTrend SSA ep looks like nice classification algorithm.

The ASCTrend sig can be used as a signal.

Let an independent reader make the tests and contribute.

3 years of back - testing is fine for me.

If you would eventually start with demo trading for the first time, day trading is the last thing I would recommend.

All the best have a nice week - end

 

We are not in Kansas any more

I think that I need to add some considerations to trend analysis. I am repeating some things from another thread in FF but it helps me not to write the same things again. This post is about the thrend and the iVAR and FGDI, and the fractal characteristics in general. I am repeating some things, again and again. Those ideas are quite new in the technical analysis and are not really fully established.

However it is not just me. Consider the article of Ehlers, Fractal Dimension As A Market Mode Sensor.

Fractal Dimension As A Market Mode Sensor - by John F. Ehlers and Ric Way

TRADERS’ TIPS - June 2010

Those ideas were present in this forum long before this article, and discussed by some forum participants, I just organized the things a little bit and I added some new hypothesis.

The trend may have different meanings depending on your method of defining it, especially in technical analysis.

The trend may be just a visual perception of a price movement over or below a MA. The trend may be just the output of some classification algorithm.

Or,

The trend may be just higher, highs and lower lows, or the trend definition may require a break - out over previous resistance in order to be established.

Some price action school of thought may have another definition of the trend.

For example make a google search: The Law of the Charts by Joe Ross and you will see a nice example. I recommend this free e-book a lot. The difference of this school of price action is that it is focused on objective price action recognition methods.

The fractal dimension is related to other characteristics, that are the fractal geometry of the price - time series. As long as I observe it, that appears to me as an independent characteristics that should be examined together with the volatility. Together they give you some information. The direction (up, down, sideways) of the price time series is another independent characteristics.

There are some patterns of the fractal dimension change, like its ciclicity and break - outs. It is possible to use it in a fuzzy way to make a market classification together with the volatility. You can use it to measure the probability of the next movement, persistent or antipersitent.

The different fractal structure underlines different characteristics of the noise you may expect. Really a lot of things the bad thing is that it is not really organised as ideas as it would be if it was in a book.

This is a different way to look at things that does not replace the technical analysis. Not at all. It is independent and from there comes its strength.

I introduced the entropy indicator too that add another level to the analysis.

The Lyapunov exponent gives another layer of analysis.

The idea is to build a complex picture of the market state.

However I recommend to read the entire threads before making any assumptions for what works or not works.

This is a job, some of the best human minds and the most powerfull computers are working on it everyday.

You are not in Kansas anymore.

I just do not recommend the combination (especially intra-day) between a lot of money, high leverage, little experience and a lot of greed.

PS

The greed come easy to me when I see such a performance.

Files:
signal_2.gif  30 kb
 

amazing

Hi John,

I want to tell you that I'm happy today.

I have put a lot of hard work in the tools you seem to understand so well and I know now that I was right in releasing them to the public.

Unfortunately I have never been able to profit from my work on spectral analisys (I mean in trading ;-) ) . Not even with the more sophisticated tools that I have developed with the help and intuitions of some real good friends of mine. They are not released to the public, so don't ask me to disclose any information.

I am convinced that all the tools avaliable (for free or not) are useless unless they help you to understand the market in a better and more profound way.

So, keep on with your good work. I wish you all the best.

Ciao

 

,Well , it's rather surprising me

Hi Richcap

IMHO, you should be successufull , because the digital filters show good exit and buy signals. Also, iVAR and Hurst difference are very good confirmation signals. So, I do not understand why you are not successfull , in your case.

I would like to know which tools are you using ? Are they come from Elite TSD ? That could explain why they are not released to the public...

If they are useless, they will be released sooner or later.

Also, which kind of Expert Advisor are you using ? which kind of SL ? (can explain a lot's of things)

Or, better, do you use IA? Are you using RapidMiner ? NeuroShell ? SVM w/ NeuroTrend ?

From my side, I am currently trying to figure out to compute Lyapunov indicator, telling if the market is chaotic or predictable. That means, is the market is chaotic, playing w/ strategies (even you win), is like gambling at the casino

In case you had some code sources somewhere...

Regards