a trading strategy based on Elliott Wave Theory - page 63
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
As for the code - no problem. The only "but" - it may take me a little longer than you expect. I consider myself an amateur, not a specialist. So don't mind me. You can send code here: yurixxx [at] gmail [dot] com.
I want to add something else for better understanding. :-)
I myself have not yet implemented algorithm of Hurst index calculation. I am busy working out another step in my Expert Advisor and I do not want to leave it halfway. Although, I studied Peters for some time to master the theory. So Hirst's calculus is on the line and it will hopefully come in a week or two. You can wait if you like.
On the subject of cyclicality, I think you are worrying in vain. Just because Hurst started with an inflow doesn't mean anything. The meaning of his index has long been understood and generalised to all random processes. In this case you can consider a cycle a period of time before a new quote appears. And the fact that these times differ does not play a role. That's the nature of this discrete process (unlike inflow, which is continuous and constant - so it has to be discretized somehow). What matters here is the series of random numbers, for which it's the consistency that counts, not the distance between them on the timescale.
There is another detail. Maybe you haven't considered it. Before calculating Log(R/S), the sample is normalized. That is, each of the numbers in the series is replaced by its difference from the sample mean. This is equivalent to approximating a horizontal line. This does not affect the value of R, but the value of Sko changes significantly.
I sent my code to your mail (if you don't get it, let me know), besides I posted it in my first post (page 30) with test results.
As for targets, there's just the thought of using it along with my digital indicators. But I need to know that I calculate exactly Hearst, I calculate correctly and have the right to apply an estimate to it as a Hearst indicator. And with the inflow - I will figure it out myself, but I also hope for the forum participants.
I think I took into account all of them. I did it strictly according to formulas.
That's exactly what I'm going to do as well. :-)
Got the letter. I will have a look at it.
That's exactly what I'm going to do too. :-)
Got the letter. I will have a look.
Finally! Found the "digital guy"!!!
Do you mind if I ask some questions by mail, regarding DSP?
For example, I can share my functions under MT, implementing forward and reverse discrete cosine transformation (from all variety I chose implementation, as in MATLAB 7.0 - absolutely exact match of forward and reverse transformation results with this package)
I want to clarify (on some philosophical level) about the perception of a cycle as a time before a new quote appears. Does it mean a period (an hour, 30 minutes, a day etc.)? I.e. we compare the time period in Forex market with the year in the book? Or we may evaluate the cyclic behavior of the data and say - stop, we have found the optimal N for stability prediction for the specified number of bars ahead.
And another question. Suppose I've decided and I want to estimate stability of the formed structure, say, for Close[] for some number of bars ahead. What do you think to take for the "inflow" in this case? Not so long ago Rosh recommended to take the delta, by the way, it approximately gives the calculation data similar to Vladislav's ones for all the bars. Although, as I found out - is considered some, from my point of view "wrong", but well working indicator, perhaps so it should be considered for small N, but again - Feder states that you can not.
No problem, any way I can ... Especially if you also explain what DSP is. :-)
Thanks for the offer. So far, no need in general.
Any mathematical tools only make sense within a particular method.
What I am trying to implement, so far, does not require cool mathematics. As it seems to me in any method the defining role is played by a complex of ideas, which underlies it (a necessary condition for success, though insufficient :-). In Vladislav's method, for instance, it is at once structural and clear in the use of the ideas of theoretical mechanics. Therefore, it is not surprising that it works so well.
That is what I am doing. I want to formulate my own approach that is sufficiently ideologically grounded.
And mathematics as needed.
No problem, whatever I can do ... Especially if you explain more about DSP. :-)
Thanks for the offer. So far there is no need in general.
Any mathematical tools only make sense within a particular method.
What I'm trying to implement so far doesn't require cool mathematics. As it seems to me in any method the defining role is played by a complex of ideas, which underlies it (a necessary condition for success, though insufficient :-). In Vladislav's method, for instance, it is at once structured and clear in the use of the ideas of theoretical mechanics. Therefore, it is not surprising that it works so well.
That is what I am doing. I want to formulate my own approach that is sufficiently ideologically grounded.
And mathematics as needed.
DSP - "Digital Signal Processing". It just seemed to me that your indicators are based on the same approaches. So I got excited, as it turned out - for nothing.
I use Fourier transform for an appropriate scheme of lowpass filtering. It works slowly but unlike my current calculation of Hurst index it gives good results. :о))) But it is OK, I will figure it out with Hurst too.
Vladislav's methodology is very interesting - hats off to him. When I wrote "I can share my functions under MT..." I wasn't trying to say what approaches and tools I use in building my trading system. The MTF is a small part of what I do now. And assuming that you also use DSP (probably reacted to "digital indicators"), I decided to offer written functions.
The appearance of a new quote is a tick. Thus I meant the time between successive ticks. Their average frequency of their appearance is about 4-5 per minute, so it is not comparable with any timeframe. Correspondingly, the period is not comparable to a year. What I meant was the following.
Each successive price value is, in general, a random number. Their appearance is not programmed in any way. It depends on the processes on the international exchange market, not time. That is why during the day (when these processes are turbulent) quotes come in 10-15 every minute, and at night (when everyone is sleeping :-) you have to wait for 2-3 minutes. So, for this random process time can be compressed and stretched. So time, as a measure of the dynamics of this process, is a poorly suited quantity. Much better is the fact of course change itself.
Note that in Peters N is not time, but the number of an item in the sample. Its relationship to time is, of course, there. But as a rule it itself is of a random nature. Hirst's year is dictated by the nature of the process - the seasonality of the seasons. Here the nature of the process is completely different. A change in the quotation can happen at any moment. Both in winter and in summer.) And what is important for you is that it has taken place, that there is a new element of a random series. And it does not matter how long you have waited for it, be it 1 second or 1 hour.
NB. This is my own attitude to the situation. I haven't seen it proposed anywhere else.
I totally agree with Rosh. Inflow is best associated with the delta and Close[] with the accumulated volume in the reservoir. However, it seems to me that since Close[] is the cumulative total from the delta, the two series are very closely related. Therefore, the Hurst coefficients for them must somehow correlate. But you should hardly rely on my opinion. All I am saying is based only on my general understanding. When I work with my hands, I will be able to form a more informed opinion.
Now I understand, what your phrase "I've found a digitizer" meant. :-)
Indeed, I use completely different approaches. Although spectral analysis of random series is, I think, a very interesting direction. However, too complicated for me. And too far from my specialty to take it up now.
I didn't think you were going to make your approaches public.
Sergey, I appreciate your offer and am very grateful for your openness. Human qualities are superior to professional ones! I refused (for now!), but only because I am afraid of being carried away by the new when the old is unfinished. After all, I tried to find more or less simple sources on the net to make sense of it. So don't get frustrated, I still hope to ask you to share these features later. :-)