Volumes, volatility and Hearst index - page 22
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
to Yurixx
Серега !!! ТщательнЕЕ надо. Я тут выделил кусочек, так это я о нем. Я получил величину показателя Херста для абсолютно, бесповоротно и окончательно случайного ряда, сгенерированного встроенным ГПСЧ. К котировочному процессу он имеет такое же отношение как я к Нобелевской премии. Это был (поклон Вита) контрольный пример. И все !
I don't want to upset you right away, but if we calculate correctly (for example, if we use Shiryaev's algorithm or other more precise algorithms), then the average result is 0.5-0.6 for large series of quotes (the process as a whole).
And when I say trend/float, you have to be sensitive to it too. It's not the first year we've been hanging out here. It's been understood for a long time that these terms are relative. So think of them as terms of fuzzy logic.
I understand, but I'm trying to persuade you to be more clear (such a moment is coming) :o) And it's already worked:
I would highlight the band around the 0.5 mark where we smoke bamboo. Above that is conventionally trending. Depending on the Hurst value we may predict the duration and / or size of the trend. Below - conditionally flat. Depending on the Hurst value we forecast the oscillation range. All this all, of course, on the basis of studies which will show the corresponding correlations. If they cannot be found, forecasting is hardly possible.
Reasonable, but is it enough? Maybe it makes sense to develop an alternative criterion for classification, e.g. - "geometric", which would unambiguously classify the state. After all Hu=0.99 does not explicitly say anything about that the next realization of the process will be very far from its mean (it still needs to be found) - it may not "go anywhere" at all. We will trade according to "geometry".
(Just in case) let's clarify the interpretation ofHu, (a kind of first approximation to the model) I propose:
Zone - some kind of boundary, characterizes, among other things, the error of calculation
(*) it should be remembered that it is the future that is of interest, the present is already visible
(**) we need to define what is the mean, or maybe the "initial conditions".
Sergei, we need him too. Where are you going to put him?
into good hands :o)
to Lea
Good evening)
I would be glad to, but I have no time for research.
Yes, we all have problems with time :o(
to Prival
I haven't been able to get him anywhere. he's kind of a dope https://www.mql5.com/ru/forum/102239/page12
It's not stupid, it's just that when you use it, you need to doR/S analysis, it's a whole methodology (and not an indicator per se), the main object of research is the dependence itself - its shape and you can't just put it on automatic. If you take it seriously, generally speaking - there is no full-fledged power dependence, it is observed only in a narrow segment and it still needs to be understood. Moving to log-log coordinates and trying to determine the degree, most often no one even looks at how linear it is, ie, whether the model is adequate, even a simple coefficient of determination no one thinks. By the way, Yuri - this should also not be forgotten.
to Yurixx
I have a "physical" question. Hirst, in 1951, published the phenomenon he found in the behaviour of the total annual runoff (denoteQ). He assumed that the runoff formation process (apparently, as a phenomenon in general) of the Nile is random, he expected this pattern
Q~k*(n)^0.5
But it turned out that:
Q~k*(n)^0.7
that's the whole effect. The picture shows this very flow of the Nile, over a couple of decades, and in 67 the Aswan Cascade was commissioned, and the flow pattern changed altogether:
Now, to be honest, I don't really understand this kind of correlation for a natural phenomenon at all. Any natural phenomenon, even the worst one, always has an energy limit. This flow cannot tend to infinity at large n, not even for an aggregated process. It may not be predictable, there may be large spikes and fluctuations of this energy and, accordingly, the runoff, but there cannot be infinity in its deviations, even if compared with early observations. Something is not right here.
And let's define again, the indicator of what exactly are we investigating? What dependence and what process: increments, R/S ratio, ... But I think it is better to go to the structural function.
Considering that two factors (rainfall and heat) influenced the runoff, harmonics are visible in this data, not Hurst. Even one factor is sufficient - the solar cycle.
It determines the cyclicality of precipitation (albeit with a shift) and temperature...
Sequoia rings need to be analysed. That's where the ticks are.
:)
Considering that two factors (rainfall and heat) influenced the runoff, harmonics are visible in this data, not Hurst. Even one factor is sufficient - the solar cycle.
It determines the cyclicality of precipitation (albeit with a shift) and temperature...
Sequoia rings need to be analysed. That's where the ticks are.
:)
I think the factors were up to ... I mean a lot. The Nile is long (flows almost all over Africa), for example, in Egypt the rains are rare, once every five years and it is not a fact. But why Hirst, looking at this graph, would suggest that their behaviour is random is a mystery to me. Have to look up his work, read and delve into it.
PS: Yep, ticks all over the place.
Reasonable, but is it enough? Maybe it makes sense to develop an alternative criterion for classification, e.g. "geometric", which would unambiguously classify the state. After all Hu=0.99 does not explicitly say anything about that the next realization of the process is very far from its mean (it still needs to be found) - it may not "go anywhere" at all. We will trade according to "geometry".
(Just in case) let's clarify the interpretation of Hu, (a kind of first approximation to the model) I propose:
I am used to perceiving Hearst value areas as follows
Therefore, a value of 0.99 clearly indicates that the process tends to continue in the current direction. Another thing is if the Hurst we have is local. Then it itself can change at any moment. Correspondingly, the forecasts will change.
I think the factors were up to ... I mean a lot. The Nile is long ( it runs almost all over Africa ), e.g. in Egypt the rains are rare, once every five years and it's not a fact. But why Hirst, looking at this graph, would suggest that their behaviour is random is a mystery to me. Have to look up his work, read and delve into it.
PS: Yep, ticks all over the place.
Do you think watering animals and peoples of Africa is also a factor?
And the rings filtered/increased naturally and proportionately...
In all sequoias. Prival is right.
You can see better from the glass.
;)
A word or two more actually about Hirst.
You may get the impression from this thread that I think this indicator is nonsense, silly, the wrong measure, or something like that. In fact it isn't. Hurst is quite an objective indicator, related to other strictly mathematical measures. This alone already suggests that it is accepted by mathematics and is an objective characteristic.
However, we should still be careful about its content.
The Hurst index is a marginal measure. It is defined as a limit, asymptote to which h is tending in the known formula for the normalized spread when the number of counts in the interval increases to infinity.
A complete analogy with the Law of Large Numbers. In the limit of LNT many theorems of probability theory and statistics are proved. In this limit even all distributions tend towards normal. So why is it that the normal distribution no longer suits us in the market. And in any field, people want to know the distribution to which the process obeys now, not in the limit of the distant future.
This is why the convergence of the process comes to the fore. If it converges quickly, then limit theorems and normal distribution can be used with good approximation at early stage of statistics gathering. If not, then, imho, all results of FFT application can be framed, hung on a wall and admired at a cup of tea. And for practice it is necessary to look for something more adequate.
The historical series of quotes is short. The market is constantly changing, both as a result of changes in the financial and economic situation and the processes that shape it, and as a result of changes in market technology, its technical support (for example, the transition from 4 to 5 digits). And the TS has to be adequate to the market all the time, not in the long term. We are all going to die in the long run - that's what some famous trader said when asked about the market situation. It's hard not to agree and dangerous not to take that into account.
That's why I think that Hearst, in its classic form, is poorly suited for use in trading. It needs either to be localized somehow, or to find other, more practical measures to estimate market behavior.
1. If you will be interested - here is the link to work that calculates the maximum drawdown, in addition it calculates the maximum spread that is proportional to the root of T. Also I attach a link to the work of Feller, who also did not feel lazy and calculated the maximal spread for SB and showed that it is proportional to the root of T.
2. In light of point 1, my assertion that Jurix's calculation confirms the hypothesis that the average run is proportional to the average spread is considered outdated and incorrect. What's there to confirm when it "turns out" to be an accurate analytical result long ago. I can now argue that Jurix's calculation doesn't confirm anything, except that the GCF he uses is correct.
3. The notion that H is an asymptote, as stated by Jurix above, does not reflect the essence of the Hurst index, nor the way it is determined. R/S analysis does not compute any asymptotes or approximations to them. R/S analysis does not use only 2 points (as Jurix does in his last formula, unfortunately it is still unknown how he does it in his program), but hundreds or thousands of points in order to estimate the Hurst exponent. Assuming that Hearst, Maldebrot and the author Peters know how to calculate asymptotes or the slope of a line by two points, the question immediately arises - why did they invent or use such a complicated method as R/S analysis to estimate Hearst? Why do they cut the series into different pieces over and over again, rescale them, recalculate and weigh them, then lay them out on the plane by the dozens and hundreds, all for the sake of determining the slope of the straight line? You couldn't figure out the slope of a straight line from two points? Idiots, the real ones. Not like some geniuses.
4. Hurst's exponent is not the limit from the formula R/S = c * n^H. Otherwise it would be counted that way, or even by the formula that Jurix suggests. R/S = c * n^H is just the right formula, behind which lies the essence of the Hurst exponent, this essence is confirmed by repeatedly checking the equality in this formula for different scales of the series under study, not by converging the series to an asymptote. Forgetting about the essence, and reducing Hearst to the asymptote in the analytical formula, we come to what Jurix came to - h = [ Log(R1/S1) - Log(R2/S2)]/[Log(N1) - Log(N2)] - wrong estimation of Hearst exponent.
5. Unfortunately, the wrong estimate of the Hearst exponent ironically fell well on SB. SB is a well-studied spherical horse in a vacuum, with which now we can do anything we want. Divide the midspan log by the time log and get 1/2, for example. And call it a Hurst. Easier to say straight away, I have the Hurst formula for SB: H = 1/2. Passes any "my" control example on any "my" row, so don't pester me. Nevertheless, I will pester you again and ask you to post the code you, Yurixx, used to calculate the Hurst figure. So far you have not confirmed that it was Hearst that you were counting. Obviously, you are afraid that not only I, but also anyone else who tries to use your program, will find you out.
to Yurixx
Я привык воспринимать области значений Херста следующим образом
- =0.5 - SB
- <0.5 - RMS grows slower than for SB. Any trend tends to reverse.
- >0.5 - RMS grows faster than for SB. Any trend tends to continue.
Therefore, the value 0.99 unambiguously indicates that the process tends to continue moving in the current direction. Another thing is if the Hurst we have is local. Then it itself can change at any moment. Correspondingly, the forecasts will change.There is a trick - if we model the whole process with the Hurst index, say, 0.9 (the model itself is not so important), we may be surprised to see 30-40% of the series not trending at all. And where is the line on uniqueness?
It is another matter if Hurst we have is local.
Then we need to introduce time dependence and this is correct.
PS: So what about my question?
to FreeLance
Do you think the watering of animals and peoples of Africa is also a factor?
And the rings filtered/increased naturally and proportionately...
In all sequoias. Prival is right.
You can see better from the glass.
Statistically, Prival is right, shall we say, not very often.
There must have been a reason why Hearst thought runoff behaviour was random, maybe he just knew more in that area, not just "rain and heat". And in general, grow your sequoia in peace, and if you have something to say - say it clearly, because it is not clear what the claims are about, the flora and fauna of Africa, the Nile or the quality of the sequoia.
Farnsworth:
There is a tricky thing here - if you model the whole process with a Hearst score of, say, 0.9 (the model itself is not that important), you may be surprised to see 30-40% of the series not trending at all. And where is the line on unambiguity?
I just don't understand the identification of persistence with a trend. It always seemed to me that consistency should rather be understood as predictability (or the same stationarity). In that sense a decent flat is no worse than a trend.
I'm only in favour of defining terms and concepts. Judging by the sharp decrease in the intensity of communication, Yuri has already started to calculate something - and again 10 pages of text will appear, bringing colleagues closer to "understanding understanding" :o)