Log in

Market State Analysis and System Trading

image

The difference between Elliotware and the Market States Analysis

The market state analysis is different from the Elliotware analysis. In the Elliotware we try to determine a constant market structure otpimize on that structure and hope that the solution we have will work out of sample. I have found such a structure on the 1h time frame on the EURUSD. Before that structure we made out of sample analysis on the uptrend with trading solutions. In those examples I used commercial systems (Trading Solutions and Neuroshell Day Trader) to make the analysis because they offered very good statistical evaluations of the trading systems so I could share the results.

On those examples as they were not rule based we used neural nets it was not possible to use robustness analysis. You know you really do not know what really does the Neural net, what really matters is what data you are going to feed to the net and how you are going to preprocess the data. The strenght of the commercial software is that it has some foolproof safeguards that avoid to the people who do not understand in neural nets to make big mistakes with the neural net parameters (As the Neural nets that are used are different Neuroshell uses some kind of cascade algorythms that adds more neurones as needed, and has a default limitation on the number of neurones limiting the possibilities of overoptimization, Trading solutions is worth because it helps you to follow the good practices for the ration between samples to weights ratio for better generalization).  

 

The Market States Analysis

However this time I would make something different as I am using a rule based system of Brain Trend expert  (you can download it here if you haven't downloaded yet). 

As it is a rule based system we are particularly interested in the optimization space and to find a global optimum.

 

image
От 08 ноември 2011

 

The Alternation of Market States

There is something very interesting we can observe on the markets and that is the alternation between the market states. This is not magic of course we can have some theories why this occurs but I do not think I know exactly why it happens. The most probable explanation is as there are phases of accumulation of market orders those orders have to be activated. The transitions between accumulation and activation from a common technical analysis perspective is a break - out. You have a trading range and you have a break - out (the mother of all  technical analysis strategies).

So basucally we do not  have a two step process

accumulation - distribution

we hace probably a 3 step proces:

accumulation - activation - distribution 

And as you can guess there is a very complex dynamics between the market states and the processes of accumulation - activation - distribution.

Well we know that most of the sharp impulses are due to cascade of running stops. There can be said a lot more. But let keep it simple right.

So take a look at this chart. 

 

image
От 08 ноември 2011

 

We had a trending and volatile market state, we had a technical break - out. What you could expect is an impulse. So what to do. It is possible to deploy a robust trading system for impuls environment. (By the way there was a change in the fractal dimension structure but I will avoid to comment this in this example I can say that we have also a fractal break - out a transition from a state of High Fractal dimension towards a state of lower fractal dimension).

And I deply on 3 different time frames the Brain Trend expert. This is not an out of sample analysis, this is the optimization space but as I optimize only 1 parameter it is not multidimensional (lol, simpler the better),

 

 

image
От 08 ноември 2011

 

image
От 08 ноември 2011

 

image
От 08 ноември 2011

The analysis of those three optimization surfaces showed that the system was robust  almost on all possible optimization parameters and that is a robust system. So you can choose a default parameter somewhere at the global optimum. 

On this optimization report it is not shown but it is good you have to make the analysis by yourself using the optimizartion tester. 

What is interesting is that the drow down is low. So you can set your risk level for example of 5 %. You will know immediately if you are right or not. Of course a visual inspection of the chart is very useful.

The advantage of playing a system in this manner over manually play the break - out is that a robust implulse following and momentum driven system will not let you go against the trend, even if the first break - out is fake, the system will know when and how to reverse.

Comments

  • JohnLast 2787 days ago

    This post was written some 39 days ago. Today I decided to make again some tests as the market states still remains quite similar to what it has been before. And again surprise surprise the same brain trend perfomed again very well at all settings since the beginning of 2012. 

    We can take a look at the 15m, 30m and 1 h time frame on EURUSD. I would say if the market state keep the same characteristics the system would be profitable. What I would not like to see is flat market or very noisy market  with high fractal dimension.

     

    image
    От 08 ноември 2011

     

    image
    От 08 ноември 2011

     

    image
    От 08 ноември 2011
  • JohnLast 2785 days ago

    Today I made some similar tests with the ASCtrend expert. The tests were on 15m time frame and 30 m time frame. The results were very similar (on 1h the expert was loosing).

    The results were very similar with the performance of the Brain Trend. Can you infer any definitive conclusions from those results. I do not think so, but we have some hypothesis. Some experts are very correlated in their performance, so when they win they win together, when they loose they loose together. 

    Here both Brain Trend and ASCtrend experts are market followers, so they would perform very well in similar market conditions. So the job is to find a robust market follower, that is because there are market follower (trending systems that are not robust). And when the market state is there you just run it, even if the settings are not exactly the most optimum, and you never know what would be the best settings you are confident that it will work. And if the markets conditions are not there it can't work. And that is it, it can't. 

    Having that in mind, the quest for the perfect expert is pointless, there is not an expert that would perform anytime. So the work would be to find a robust systems for particular market state. And the processi is to switch between the systems, but that requires a critical set of meta-knowledge.

    In this example if you see that ASCTrend or Brain Trend do not work, it does not mean necessarily that they are worthless, it means that the actual market state is not within the performance parameters of the system.

     

    image
    От 08 ноември 2011

     

    image
    От 08 ноември 2011
  • JohnLast 2727 days ago

    Very interesting film, definetely for the watch list.