a trading strategy based on Elliott Wave Theory - page 286
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Seryoga, there is no contradiction. I've written quite clearly that predicting a pattern is utopia. In other words, I'm not predicting that now, from the next bar it will form, for example, "the head and shoulders" with such parameters.
I'm only hinting that an attempt to "recognize" patterns by their conversion coefficients is a waste of time (in my humble opinion). My job is to warn.
PS: Serega, special "thanks" for "Fundamentals of stochastic financial mathematics". :о)
Suggest where you can drop 9 metres of manoscript.
Those are details, but I'm interested in a general question: is it possible to predict an incipient pattern using this method? What's the tricky part of the method. You, as an expert in this field, point out a possible direction of search.
Yes... Dear Neutron, you certainly understand that you have asked a global question. A good question. Predicting an incipient pattern is like predicting future price movement. Whether it can be done and with what probability depends in principle not on the method used, but on the properties of the market. You have shown it yourself perfectly for statistical methods. Wavelets are just one of many methods and they are probably no worse and no better than some others. It's just that, indeed, I know this topic quite well and for me it is convenient to go this way. But I don't expect miracles.
If the market is efficient, i.e. completely random, then it's a casino. Then what to do in it? No method will help. Something tells me, however, that all is not so bad - in particular your results and Pastukhov's thesis and other information. Or maybe I am just an optimist...?
I will try to tell coherently the approach that I am going to test, and you decide what to do with it.
Looking at the results of wavelet transform of price charts you can see almost regular structures that appear, evolve and disappear quite regularly. So far, it's just a feeling, an impression of what is visible to the eye. No more than that.
Some structures can be associated specifically with rising or falling trends.
Further. You yourself suggested that there are periods of deterministic and stochastic movement in the market. I like this idea too. You can separate one from the other with statistical indicators (correlation, Hearst, etc.). Assume that arbitrability is possible on deterministic plots.
What I am going to do next. I am going to write a block that recognizes meaningful market structures (trends, etc.) based on wavelet methods. It could be called pattern recognition, but I don't feel like it because this notion has no unambiguous interpretation. For some people a pattern is a standard market pattern (two tops, head and shoulders, etc.), for others it is a combination of candlesticks, etc.
Then we run it all through the history in order to find characteristics such as average duration of structures at different scales, frequency of their appearance, dispersion of these values and others. And we do it at deterministic portions of the price movement. At this stage we already have figures in our hands, i.e. objective information. Further, if we can determine the current state of the market - whether and for how long the previous movement is going on, or whether it is close to change, or whether it has recently changed, etc., using recognition of the current market image and also statistical data on the nearest history, we can make a reasonable decision (to open or not to open a position now and in which direction), getting a statistical advantage. What would that be? I do not know yet. Unfortunately, this is the only answer to your question. If it is 1% it is clear that there is no need to bother with it, if it is 10% - it is another matter - then it is a matter of technique.
That, very schematically, is the idea and my plan for the near future. The task is not an easy one - I can already see many pitfalls, but it is manageable. It will also take a lot of time. I cannot yet say how long. As soon as I get specific and reliable data I will share them. I am still working and at the beginning of my way.
PS. You said a few posts earlier that you have a great book. You promised to publish it. I would like to read it. Is it possible?
You're probably right. I got it somehow easily and completely by accident. I was fascinated by the fact that it is very simple and unambiguous. If at least one person can make use of it - I'll be glad, but if not - it must be destiny. I have absolutely no pretensions to anything here!
I have no pretensions. The procedure is really interesting and simple, it's a pity that the analysis of its properties is very difficult.
P.S. Thank you for great material on statistical properties of random and price series! I read it with great pleasure. I liked it very much and found it useful.
Not at all, if it was useful and understandable.
PS. You said a few posts earlier that you have a great book. You promised to make it available. I would like to read it. Is it possible?
In my opinion, which is not supported by anything but intuition, the market should be very simple in its reaction to the story. I say this in the sense that it is impossible for it to implement deterministic complex structures like patterns. Otherwise, we have to accept the existence of a long-remembered market memory, the realization of which requires a stable and strong relationship between players (collective phenomena).
As for the manuscript, suggest a place to put it.
http://thenorthenwind.googlepages.com/--1.djvu
Shiryaev A.S. - Fundamentals of Stochastic Financial Mathematics - Volume2.djvu
http://thenorthenwind.googlepages.com/--2.djvu
Shiryaev AH - Mathematical Formalization of Japanese Candlesticks.rar
http://thenorthenwind.googlepages.com/-.rar
Will keep for a week.
I would be grateful.
[later] applied to disserr.ru, for more details, wait to hear back.
[came back]
You have submitted an application for information on the thesis on the topic:
Statistical forecasting for the construction of effective trading strategies in the foreign exchange market
Year of protection: 2003
This work is protected in Russia
Unfortunately, the work plan on the requested topic at the moment we can not provide, because the work is in the archive of the institution.
But if you are seriously interested in receiving this scientific work, write to us (info@disserr.ru), and we will send a letter to the applicant.
You yourself suggested that there are periods of deterministic and stochastic movement in the market. I like that idea too. You can separate one from the other with statistical indicators (correlation, Hearst, etc). Assume that arbitrability is possible on deterministic plots.
...
Apparently this thought still stops many of us from leaving this market. Just here lies a lot of tricks. The nature of Hurst index is such that according to the classical calculation schemes (including the method described by Shiryaev) its values for the price series will vary in the range of 0.55 - 0.9. Moreover, 0.55 is a very rare value, 0.7 on the average. It seems to be OK - the memory is always relatively long, but how can it help you to separate flies from cutlets? :o)
...
What I am going to do next. To write a block of recognition of meaningful market structures (trends etc.) based on wavelet methods. I could call it pattern recognition, but I don't feel like it because this notion has no clear cut meaning. For some people a pattern is a standard market figure (two tops, head and shoulders etc), for others it is a combination of candlesticks etc.
...
Let me give you a bit of philosophy. The success of wavelet analysis application in the fields of human activity you mentioned earlier, including my favourite aircraft engines, owes much to the possibility of obtaining a "benchmark". Well, for example, the vibration of the same engine without a defect in the blades, in medicine - EGC of a healthy person. It can be compared with anything and determine absolutely precisely what kind of problems a particular person or engine has. I suppose that's where your problems will start, namely the inability to get a 'benchmark'. It cannot be unambiguously identified on price charts, and certainly not on information-overloaded pretty pictures (excuse the word - "muzilki").
My modest experiments (of course, I cannot pretend to be 100% reliable) showed that in fact BPs taken separately "by eye" and symbolizing different structures (trends) have absolutely identical characteristics making it impossible to distinguish them objectively from each other. And I was quite sad...
to Neutron
Sergey, sorry! - Totally slipped my mind.
Suggest where to throw 9 m of manoscript.
Don't worry, I also suffer from forgetfulness, .... forgot what the word is called. :о))))))
As for the manuscript, suggest a place to put it.
It seems to me too that the market has no long term memory. But on the other hand it seems (also pure intuition) that the market behaves similarly in response to similar influences. Who is influencing it and how it is influencing it is behind the curtain and we don't know. More often than not, the reaction is likely to be ambiguous and very indirect. But it may still be possible to catch it. Sometimes, however, at the moments of abrupt and strong price movements one can observe the textbook picture of "forced oscillations of the resonator with losses in the presence of noise". These considerations somehow support the positive "mathematical expectation" of my hopes.
About the manuscript. Thanks. Already downloaded it from the sources suggested by Northwind.