Log in

Market State Analysis

By JohnLast 3545 days ago Comments (5)
От 21 октомври 2011

Here I would like to show an example of the multidimensional market state analysis. I use 5 m time on EURUSD. The methodologiy of Curtis Faith is explained here.

It appears like the perfect trend but inside it we can find some different market states. 

1. Trending and quiet 

2. Trending and volatile

3. Ranging and volatile.

What is interesting here is how the transition operates. The transition between trending and quiet and trending and volatile comes after a short horizontal consolidation.

The second transition between Trending and volatile also came after a short consolidation.

What we use here is visual pattern recognition between the market states. What we do is that we want our system to operate successfully in some completely different market states. And very often our systems do so. In fact sometimes we are very ingratefull because we are getting good results often in market states that are beyond the operating desing of our systems.  


От 21 октомври 2011

Here on this shot I add the iVAR (Variation Index estimation of Hurst exponent using small data set). With that information we can spot the regime shift somewhat better than using the visual pattern recogniton. We see that there are two major regimes, of high fractal dimension and low fractal dimension (the iVAR index is somewhat different than the fractal dimension indicators in its readings because the center line is not 1.5 but 0.5).



  • jaguar1637 3475 days ago


    There is a good new article regarding analyzing the Time Series


    Please, take a look on this


  • JohnLast 3475 days ago

    This is one of the major achieements for mt5. Really great. For the first time I see Arima for metatrader. 

    However arises the question, how are we going to use those tools? 

    I would like to cite a part of brilliant post that I share:

    "The market is not a gaussian structure, most things are not and variations in behavior which does not correspond to the normal gaussian distribution will be found if the analysis is large enough. Indeed the gaussian distribution curve is just one instance of a family of distributions curves which blend into one another. The dominantl distribution curve of the market at any one time changes in different directions under the stimulus of news, heavy buying or selling, etc.. Usually the intrinsic change is not marked but it does exist.

    Whenever we use standard deviations, regression, etc... we introduce an error into the calculations because these measures will by necessity change with the dominant distribution curve of the market at any given time. The market moves according to constantly variable rules if you wish (perhaps rules is not the correct word), the degree of intrinsic variability may not be major but it is there."  This is Oscar from linkedin.  

    And then another guy Guy. 

    "The markets morph from one structure to the next without having the courtesy of telling us (in advance) that they will change into and with what probability. And this makes it difficult to design short term predictive trading tools."

    And that challenges the whole conception of market state analysis. We can profit from the market states provided they stay long enough.


    Here I would like to cite, Neuroshell manual for Neural networks:


    "No model (Net or any other kind of model) can give you a probability that you should buy or sell. All they can do is give you a measure of how the current condition matches past conditions in which a buy or sell was appropriate. The markets could change, or your inputs may not be as predictive as you think they are."

    So I would like to look at those tools as meta tools, helping us eventually with the determination of the market state. This is a kind of metaknowledge that is critical.


  • JohnLast 3471 days ago

    Here I would like to add another example how the market states switch from one type to the other. 

    Look how the market states swith from trending and quiet into trending and volatile. And how within the trending and volatile patterns there are triggers of trending and quiet patterns. 

    For me it is really interesting because those kind of analytical treshold cannot be achieved using the classical technical analysis. Look there are still channels but the analytical overlay is drastically different. 

    For example for the last uptrend of the EURUSD the trend begins by trending and quiet, and thet it becomes trending and volatile. So if you train your neural net using the trending and quiet period that may not end into good results for the trending and volatily. That is why it is so difficult, the market as a chameleon changes its colours.


    От Technical Analysis
  • JohnLast 3471 days ago

    I think if you look at this shot you would understand why the taks of optimizing of trading strategies is so hard. 

    For example now I had a pattern bet. The market can ramain in the same market state or change its market states. 

    If the market changes its market states one of the most probable way to do is to make a sharp break - out to new higgs by trending and quiet way. That is why my stop is so tight.

    The other way is that the market remains in the same market state being trending and volatile or eange for a while by being ranging and volatile.

    There  is a problem that the break - out may be fake and there is a broading triangle that is forming, but that is a risk, no way to avoid that risk. 


  • JohnLast 3420 days ago

    It is an open debate if the smoothing of the iVAR, or the so called iVAR smooth indicator would help.  I am a little bit sceptic, the idea of the iVAR may be contradictory with the smoothing paradigm. And of course if you enlarge the window horizon the iVAR will get smoother, no need to smooth it. 

    Anyway the debate is open. Will the smoothing of the iVAR will help you or not?