a trading strategy based on Elliott Wave Theory - page 278
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
Sergey, could you please give me a link to this calculation. I'm interested not so much in Hurst (I've already sorted it out), as in the practice of wavelet transform. As you know, I've been interested in it lately. Unfortunately, I must admit that I have made very little progress so far. Basically everything is clear in theory. But how and what to calculate practically is a total mess. :-(
I highly recommend putting it in if you want to practically study the subject. I ended up being disappointed with the predictive wavelet. The max I got was this:""trading strategy based on Elliot's Wave Theory" post "grasn 18.04.07 20:03". Those were the best approximations.
I carefully read the entire thread, although it took a long time. I got a lot of fun, and most importantly useful ideas and facts. Especially concerning statistical methods applied to price series. This is exactly what I've been missing! Huge respect to all the participants in the discussion!
I'm digging on the other side. This is wavelet analysis, attempts of applying methods of image processing and pattern recognition, fuzzy logic.
I'm very recently on FOREX, so to say, first approach to boom, - but I have great experience in these topics in another domain (own developments, external libraries and source codes, a lot of literature found in Internet, etc.).
I tried something to apply to price series (mainly wavelets). The results, although very preliminary, seemed to me interesting. There are also many ideas, not yet tested. If you want, you may discuss them.
Good luck to all and passing trends!
PS. It seems that people in this thread already tired. If I'm wrong - welcome...
You try to share your thoughts and suggestions here. The rest will be decided by the readers of this forum. Maybe this will be interesting and you will get some questions and suggestions. Although it may take some time before we get responses since not everyone has experimented with wavelets yet. But in any case it all depends on what you want to tell.
That's quite interesting for me. Maybe you can share your experience of applying wavelets ?
We all know or guess that the market is time-varying, i.e. "something" changes over time. I will explain, for example, the "trend" seems to be moving and prices follow the "trend", but some general characteristics have changed imperceptibly and strong movements can be expected.
So, I derived a formula for the price at the nth readout based on the market fractality with some coefficients. The algorithm is very simple:
(1) From the current datum, starting from some minimum, I sequentially take historical channels (or time series)
(2) For each such channel, perform
- Calculate the Hurst index (by my own formula, but I want to remind you that this is not the formula given above, but I already wrote about it)
- calculation of skyling index
- a wavelet transform, used to calculate missing coefficients for the formula
(3) Calculating the price itself.
The construction of the most probable trajectory is as follows: minimum channels calculate the price for the closest samples from the current one, respectively long channels calculate the price for the "far samples". In other words, "one channel" is "one price". It is important to note that some channels may not be taken into account due to the Hearst and Skyline values.
The ultimate goal is not to build a relatively accurate trajectory, but to evaluate the reversal zones and compare them to the same Murray levels.
That is all there is to it. I gave the results:"trading strategy based on Elliot's Wave Theory" post "18.04.07 20:03". The only thing I need is to bring the model to more or less normal applicability level - I'm lacking a scientific institute :o(
Actually I wanted Andre69 to share practical aspects of the application. That is, from what considerations a wavelet-forming function is chosen, how decomposition coefficients are calculated, what is done with it afterwards, etc. But what you have written is also very interesting. Finally, I understand how a forecast series of future values appears. I've been asking that question ever since you published your first forecast. :-)
All in all, it all looks pretty good. Even the most vulnerable place of the model is forecasting nearest future by minimal channels and farthest future by maximal channels. But there's some logic in it too !
So kudos to you. Which of these is taking you so long to calculate ?
And for that matter, can you say something on the subject ? :-)
Right, I'm not Andre69. Really, why did I get involved? It's all out of thirst for communication.
Which of these do you take so long to consider ?
That's not the model that counts for a long time. I have as many as 3 models at the moment and I don't consider it to be a model at all as the results are unimportant and I still have much to do, for example to hire a scientific research institute to modify the idea.
And for that matter, can you say something on the subject ? :-
should read, "while you're at it..." :о)))))) :
... what are the considerations for selecting a wavelet-forming function, how decomposition coefficients are calculated, what is done with it afterwards, etc.
I should note that, having reread the material, I haven't found any threads yet, except a thread about wavelets in general. For my purposes I used Morlet wavelet (I know it's not a wavelet from mathematical point of view), I haven't experimented with the others. Its properties suit me fine for the task at hand. The question "how decomposition coefficients are calculated " I don't really understand. Not a very big expert in wavelet analysis, but I had no problem with it.
"what is done with it afterwards, etc." - eventually the aggregate coefficients for the price calculation formula are calculated after "etc". I forgot to add a very important point - for each historical channel is performed:
(1) Estimating its lifetime after the current countdown.
(2) Identification of a stable "structure" that will live!!!
It is on this maximum predicted channel length that the future price is calculated. Difficulties arise if there is more than one forecasted price (i.e. several historical channels of different lengths, have the same lifetime estimate). The difficulty has so far been eliminated by selecting the average value from the suggested ones. So, these coefficients estimate:
(1) The time "not to go beyond" the boundaries of the channel.
(2) What will not go beyond the borders of the channel
Ultimately, the essence of my coefficients is to evaluate stability of the structure (since I got in anyway, let me remind you of the pictures) and eventually to assemble these pieces into a whole:
This structure can be seen if you look closely:
Or here, maybe not explicitly, but it appears:
PS: OK, I will not interfere any more.