Discussion of article "Grokking market "memory" through differentiation and entropy analysis" - page 6

 
Aleksei Stepanenko:

The waves go round and round ;) :)

Here is the information provoked perturbation, and the price went. Is it possible that when the "memory" is gradually stopped, the price stops at any arbitrary place? Or these places are still limited areas, which we consider as levels. When one person thinks, "I'll get to this point and take it." And the other one thinks: "I can't take it anymore, I will close another 20 points".

I.e. price increments are different at different parts of the price scale.

It's a question of optimisation. Anything is possible

 
Alexander_K:

Ahhhh.... First person who understands what to do with ticks. Exactly - work with ticks with sampling rate of 1 sec.

Alexander, I understand this is for scalping?

 
Maxim Dmitrievsky:

It's a matter of optimisation. Anything is possible

I meant to say, it looks like levels influence and eventually stop the price. And it is supposedly desirable to take that into account too.
 
Aleksei Stepanenko:

Alexander, I take it this is for scalping?

More precisely - when pipsing.

Intrade requires more serious thinning - up to 1 value in 10-15 seconds.

It is important to understand another thing - at any type of thinning of a tick series, a certain time dependence of the sample volume and time should be preserved. I.e. conditionally 1000 thinned events during the day should correspond to a trading session, and at night - to a day if such a flow is preserved.

It is possible, in principle, to work with linear time, i.e. OPEN/CLOSE M1, M5, ..., then the cycles are obvious - trading session, day, week, ..., but within a day the accuracy is lost when the sample volume decreases and we can only talk about trading within a week, month, ...., i.e. 1-2 trades per week. An ordinary trader can not be satisfied with this - he wants to throw up and throw down during the day and works with ticks, but he loses in the time space and fails :))))))

 
Alexander_K:

About the cycles...

Within a day, they are clearly visible. See GBPUSD charts for this month:

On the bottom indicator of the Sorcerer (currently warming up, probably on the edge of the forest) this periodicity is present in the process variance (red and blue lines) - almost a sine wave.

It is important to be able to predict the behaviour of dispersion (volatility) and make deals only when the price (or the sum of increments) goes beyond the dispersion, when the inflection point is passed (marked with green circles on the charts).

Why didn't they build volatility into prices right away? I mean, the chart should be further transformed taking it into account.

thin out the volatility ticks, for example

 
Maxim Dmitrievsky:

Why didn't they build volatility into prices right away? I mean, the chart should be transformed with it in mind.

to thin out the volatility ticks, for example.

Erm... I don't know how to do that.

I found something in time cycles of volatility (but my cycles are different from Gann cycles for some reason), I use it slowly and that's all.

Of course, I do not have the Grail, but some +20-25% are already in the 3rd month, and now I am already using it on the real. Perhaps, I will pour everything down the toilet, of course.... But without these cycles, I would have drained everything much earlier.

 
Alexander_K:

Erm. I don't know how.

I found something in volatility time cycles (but my cycles are different from Gann's for some reason), I use it slowly and that's all.

Of course, I do not have the Grail, but some +20-25% are already in the 3rd month, and now I am already using it on the real. Perhaps, I will pour everything down the toilet, of course.... But without these cycles, I would have drained everything much earlier.

I already wrote in the topic of MO that in idea it is done in one go with the help of Lambert's inverse transformation.

but it's too complicated for me https://www.hindawi.com/journals/tswj/2015/909231/.

Although there are packages for R and Py

The Lambert Way to Gaussianize Heavy-Tailed Data with the Inverse of Tukey’s h Transformation as a Special Case
The Lambert Way to Gaussianize Heavy-Tailed Data with the Inverse of Tukey’s h Transformation as a Special Case
  • Hindawi
  • www.hindawi.com
I present a parametric, bijective transformation to generate heavy tail versions of arbitrary random variables. The tail behavior of this heavy tail Lambert random variable depends on a tail parameter : for , , for has heavier tails than . For being Gaussian it reduces to Tukey’s distribution. The Lambert W function provides an explicit inverse...
 

Maxim Dmitrievsky

A lexander_K

For me cycles are still an unexplored direction. I see that there are cycles and the oscillation period changes. A lot of questions. Thanks for interesting information.

 
Aleksei Stepanenko:

Maxim Dmitrievsky

A lexander_K

For me cycles are still an unexplored direction. I see that there are cycles and the oscillation period changes. Lots of questions. Thanks for interesting information.

:))) There is certainly cyclicality in the market. But, not explicitly in the price, but in its dispersion (volatility). And the single beginning is the increments that cause both price and its variance. It looks like this - the probability density of increments as if periodically shrinks/expands during a day. Actually, if we talk about memory as a consequence of a process, this thing - cyclicity - is memory.

If it is possible to represent the price as a dispersion in sinusoidal form, well, it will be something :)))))

For reference:

Process variance is calculated by formulae, for example from:

https://en.wikipedia.org/wiki/Variance_gamma_process

Variance gamma process - Wikipedia
Variance gamma process - Wikipedia
  • en.wikipedia.org
There are several representations of the VG process that relate it to other processes. It can for example be written as a Brownian motion with drift subjected to a random time change which follows a gamma process (equivalently one finds in literature the notation Γ ( t ; γ = 1 / ν , λ = 1 / ν ) {\displaystyle \Gamma (t;\gamma =1/\nu...
 
Alexander_K:

:))) Certainly, there is cyclicality in the market. But, not explicitly in the price, but in its dispersion (volatility). And the single beginning is the increments that determine both the price and its variance. It looks like this - the probability density of increments as if periodically shrinks/expands during a day. Actually, if we talk about memory as a consequence of a process, this thing - cyclicity - is memory.

If we manage to represent the price as a dispersion in sinusoidal form, well, it will be something :))))))

For reference:

The variance of a process is calculated using formulas such as:

https://en.wikipedia.org/wiki/Variance_gamma_process

Volatility can be obtained from the transformations in my article like two fingers. the question is actually in further transformations of the initial series, namely how to sample it.

you need a reserch or ready formulas