a trading strategy based on Elliott Wave Theory - page 281
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
Don't take it as an instruction, just clarifying.
Best of luck and good luck with the trend!
What I meant was that Morlet's wavelet is not strictly a wavelet, as the average value of this function is different from zero, but I was quite happy with its properties. As I wrote, perhaps I'll tell you more about it (morally not ready yet). Moral side is very simple - what is the attitude to such a prediction? (watch, eurusd) Is it correct or not?
If I understand that it is bullshit, I will tell you in details. If I understand that there is a perspective, I will tell you, but not everything. In a way, I'm taking an example from Candid. :о))))
I've given up data filtering a long time ago, including wavelet filtering (note for Solandr that there is no technical problem with real time filtering as such, there is a problem with trading :o). To get at least some usefulness out of this mess, I need to build adaptive filtering, and that, at least for me, is the problem.
Exactly for the reasons described by Neutron I've concentrated on things with predictive properties, including Hurst and started looking for a slightly different use for wavelets (not for filters at all), about which I wrote briefly.
PS: thanks separately for the following trend. :о)))
to Neutron
To quote the author of http://monetarism.ru/article.pl?sid=05/03/13/0625201&mode=flat, I note that the folio is indeed excellent! I have 2 volumes of this work in DjVu format, 4 meters each, if the public is interested - I can lay them out.
Sure, I am ready to download it. :о))))
And, besides, I was interested in how to apply wavelets in principle, not how to apply them to forex. I have an object for research and I have chosen a tool. I just do not know how to use it. :-))
And what is this tool, if not a trade secret? By the way, I recommend to pay attention to skeletons, they are useful, at least I calculate my coefficients on their basis.
WAVELETS !!! :-)))
[quote]And...?
Looking at this picture we can talk about a particular method of interpolation of a non-equidistant numeric series (obtained from EUR/USD time series by one method or another) by linear or quadratic polynomials.
[quote]
I don't understand why you think the series is non-equidistant? The time scale is uniform. Almost... Or are you referring to weekends?
Note that TC implemented using LPI does not give any statistical advantage over DC in today's market.
I agree with you on the first sentence completely. Indeed, we need extrapolation and only extrapolation, but it is not easy to achieve! If it were simple - there would be no problem!
As for wavelet methods, I'm not saying that this is a panacea or a new grail. Not in the least! And I don't encourage anyone to rush in this direction without a second thought. For me it is simply an understandable, convenient and accessible tool for market analysis. Just like statistical methods probably are for you. All in all, it's a matter of personal preference. One more thing. I am fully aware of the value and validity of statistical methods and I will definitely use them when developing my TS, for example, well... For example for detecting periods of market arbitrage. Thanks you for interesting posts on this topic!
Let's return to wavelet methods - in fact they are the same filters, or rather a set of sub-band filters organized in a certain way. There will be phase delay, of course. Of course, there will be phase delay. Unfortunately, there is no perfect filter in nature (due to the principle of causality) - there are good and bad ones. Filter phase delay is equal to half the filter core length. (I hesitated a bit here... Well maybe not equal, but proportional for sure) This means that a short simple filter has an advantage in this sense. Wavelet filter kernel sizes start at 2 (Haar wavelet). The ones I used are 5 and 8. Do wavelets give an advantage in this sense? I don't know yet. I have to compare specific implementations. As for the perfect filter... - There are no such filters, and, say, the Butterworth filter certainly does not hold that title. I used it long ago. I don't remember its kernel size now, but it's definitely bigger than 2. Compare it to a wavelet filter.
There is also another way of wavelet decomposition, which I haven't mentioned yet, which is the interval wavelets and lifting algorithm. It is remarkable in that it does not require any assumptions about the behaviour of the function outside the decomposition interval. I haven't tried it yet. Perhaps we can achieve minimal "phase delays" here. Although the term "phase delay" itself is not very correct for this thing.
10% per month it is with spreads and on 2 month history, i.e. the sample is not reliable. For the purpose of getting statistics the real account will be opened.
Thanks for the reply.
Good luck and hit the trends!
И, кроме того, меня интересовало как применять вейвлеты в принципе, а не как применять их для работы на форексе. Объект для исследования у меня есть и инструмент я выбрал. Вот только не знаю как им пользоваться. :-))
А что за инструмент, если не коммерческая тайна? Кстати, рекомендую обратить внимание на скелетоны, полезная штука, по крайне мере свои коэффициенты я вычисляю на их основе.
WAVELETS !!! :-)))
I just didn't quite get the word "INSTRUMENT" right. :о)
Do you know an indicator that reliably bends in the right direction ahead of a future price move? Then it's the Grail!
No, I don't. Such an indicator cannot exist even theoretically. However, since we are speaking about using wavelets, I just want to note that they don't seem to offer any significant advantages over other methods of representation. And it is unlikely that any strategy can be based on wavelets alone.
А что Вы знаете индикатор, который достоверно загибается в правильную сторону раньше будущего хода цены? Тогда это Грааль!
No, I don't. Such an indicator cannot exist even theoretically. But since we are talking about using wavelets, I would just like to note that they do not seem to offer any significant advantages over other methods of representing information. And it is unlikely that any strategy can be based on wavelets alone.
You are absolutely right that wavelets alone are not enough for building a TS. I'm not going to do it that way. However I am sure that they will be very useful as a market analysis tool. It's just that so far, it seems to me, no one has seriously tackled this topic, so they don't provide anything. So far... What share wavelets will take in the TS I'm designing now, I don't know yet. 70 or 10% - what difference does it make - as long as it's useful for profit.
As for advantages in the way market information is represented, I don't agree with you. They do. You are using multiple price charts on different timeframes when trading.
That is, maybe unknowingly, but you are doing a multiscale analysis. And the main essence of wavelets is not in the details of implementation and algorithms, but precisely in their multiscale. And underneath this fact, I assure you, lies a powerful philosophical idea. If wavelets have been used with great success in designing aircraft engines, processing astronomical photographs, in medical diagnostics - I know these examples just fine - and countless others in a variety of fields, then why would they break in the market? I see it differently.
Respectfully.
Good luck and happy trends!
This is what I find especially interesting. But so far you haven't said much about it. I hope very much that it is only for the time being and that it will be continued. :-)
You mentioned that you have collected a lot of various information on wavelets. Could you post something here for your discretion ? Polikar's "Introduction to wavelet transform", Dobeshi's "10 lectures on wavelets", Vorobiev-Gribunin's "Theory and practice of wavelet transform" and some other smaller stuff I have. I'm reading slowly Dobeshi.
The problem is that there is too much theory, which I understand at my, elementary level, but cannot practically do anything. That's why I need something more or less simple and task-oriented, from which I can understand schemes and algorithms of concrete actions.
It is desirable that it is not DSP. I have nothing against DSP and well understand that any time series, including series of quotes, is a signal and can be investigated by DSP methods. However I am very far from this area and I am sinking in the terminology, jargon and terms accepted by specialists.
I was focusing on the kinks in the linear polynomial, they are not equidistant. I'm probably wrong though - after all, a node can be on a line connecting nodes adjacent to it.
It is clear that BP decreases as sampling window narrows, but the smoothing properties of the operator get worse. We have to find a compromise between smoothing quality and lag. That is why it is correct to compare smoothing characteristics of operators at identical or close parameters of their AFR (evenness in passband, cut-off slope). In this regard, Butterworth filter has minimum (not zero!) bandwidth, which significantly increases at cut-off frequency. It is in this light that it is interesting to compare wavelet-based and classical filtering methods.
If we are going to extrapolate something somewhere, there will inevitably be FZ. Indeed, sitting at the right end of the time series and extrapolating one step ahead, we get the probable value of the series in question. In the next countdown, compare the value with the true value and remember the resulting error. Repeat this procedure once more, taking into account the update of the input data for the second point, and so on and so forth. As a result we have two time series - initial and forecast. Obviously, they do not coincide exactly, but also do not diverge strongly, only shifted relative to each other by FZ! So I think the term FZ is appropriate for this case.
Now, colleagues, critique me.
I argue that any extrapolation implies that a time series (TP) has the property of "following" the chosen direction. Indeed, by extrapolating one step ahead by a polynomial of nth degree, we assume the NEED for the first derivative, the second... n-1 of the original series, at least at this step... Do you see where I'm going with this? Quasi-continuity of the first derivative is nothing but a positive autocorrelation coefficient (AC) of BP at the selected timeframe (TF). It is known to be pointless to apply extrapolation to Brownian-type BPs. Why? Because the CA of such series is identically equal to zero! But, there are GRs with negative QA... It is simply incorrect to extrapolate to them (if I'm right) - the price is likely to go in the opposite direction from the predicted direction.
And for starters: Almost all Forex VRs have a negative autocorrelation function (this is a function constructed from the KA for all possible TFs) - this is a medical fact! The exceptions are some currency instruments on small timeframes, and yes Sberbank and EU RAO stocks on weekly TFs. This, in particular, explains the unsuitability in the modern market of the TS based on the exploitation of moving averages - the same attempt to extrapolate.
If I am not mistaken, wavelets a priori appear in the area where they cannot correctly perform their functions.
You mentioned that you have accumulated all sorts of information on wavelets. Could you post something here for your discretion ? Polikar's "Introduction to wavelet transform", Dobeshi's "10 lectures on wavelets", Vorobiev-Gribunin's "Theory and practice of wavelet transform" and some other smaller stuff I have. I'm reading slowly Dobeshi.
The problem is that there is too much theory, which I understand at my, elementary level, but cannot practically do anything. That's why I need something more or less simple and task-oriented, from which I can understand schemes and algorithms of concrete actions.
It is desirable that it is not DSP. I have nothing against DSP and well understand that any time series, including series of quotes, is a signal and can be investigated by DSP methods. However, I am very far from this field and I am sinking like in a swamp in the terminology, jargon, and terms accepted by specialists.
There will be a sequel. I am preparing it. As always, lack of time. Maybe I'll post it today.
About the information. Already said that there are several pdf files with review articles. A couple of them seem to be Gribunin's translations and are quite famous. You probably have some. The others are more serious.
It would be more convenient for me to send them to you by e-mail. Mine is andre69 [at] land [dot] ru.
I have the information on the lifting algorithm only in English. The original articles by the authors of the method and their followers. If you are not confused, I can pick something up.
About Dobeshi. You are a giant! I only had patience for half the book. The maths is good, of course, but it's a long way from practice. You should only take global ideas from there.
Remark on DSP. DSP and wavelets are quite strongly connected to each other. Unfortunately or fortunately I don't know.
Regards.
Good luck!
There is certainly some common sense in this. But there is also some "but".
If the extrapolation has the property of monotonicity, its value is very low indeed. MA can only provide such extrapolation, which is why it is not used for this purpose.
But if we take something more complicated, a polynomial of degree 2, for example, it is not quite so.
Let me clarify: we are talking about extrapolation to the nearest future.
So, with a simple quadratic function (assuming that the number series really allows it by nature) you can predict the approximation of the turning point. And that's exactly what everyone needs. Especially polynomials of higher powers. So extrapolation almost always preserves direction. But it almost changes the whole picture.
And as for CA, it is, as rightly noted, dependent on the chosen TF. This reflects the fact that the series under study is piecewise monotonic in one way or another. What difference does it make whether one chooses a TF for which KA enables one to make some decisions or whether one chooses an interpolation method which can provide a relatively reliable extrapolation to the near future ?