Discussion of article "Grokking market "memory" through differentiation and entropy analysis" - page 7

New comment
 
Alexander_K:

For reference:

I'll read the whole thing, thanks. And the links from this article too.

It is interesting, if you plot random increments (well, pseudo-random of course), then visually it is not distinguishable from real graphs. However, their density will also vary randomly. And it seems to be of no use. Although it looks the same :)

Randomness

added Excel file with working example

 
Aleksei Stepanenko:


It is interesting, if you plot random increments (well, pseudo-random of course), then visually it is not distinguishable from real graphs. However, their density will also vary randomly. And it seems to be of no use. Although it looks the same :)

I checked one of the sequences:

At the end I don't know what you have there - some failure, I haven't analysed it.

But, in general, it is a regular Gaussian random process, without any periodicity in variance and it is quite problematic to win on it, but you can.

The difference with the real BP is simply colossal.

 

Very interesting, and the joint - a couple of rows at the end of the table should be deleted. I've made the changes.

Alexander_K:

The difference with the real BP is enormous.

And build an indicator on this difference :)

 
Aleksei Stepanenko:

And build an indicator on this difference :)

Build it :)))))

If only it were so easy, some people here have been struggling with this task for 15(!!!) years.

Someone finds it, of course, and immediately jumps off this forum.

 
Alexander_K:

15(!!!) years they have been struggling with this task. Someone finds it, of course, and immediately jumps off this forum.

Hello, nirvana!

 
Aleksei Stepanenko:

I'll read it all, thanks. And the links from this article too.

It is interesting, if you plot random increments (well, pseudo-random of course), then visually it is not distinguishable from real graphs. However, their density will also vary randomly. And it seems to be of no use. Although it looks the same :)

added an Excel file with a working example

And if you load this chart into mt5 as minutes and switch on time frame m15 or n1, you will not be able to distinguish it from the market chart by eye, I checked it.
 
Maxim Romanov:
you can't tell the difference from the market at all by eye.

Thank you, that's great!

But there are differences in the graphs. The Excel formula produces random increments, and the frequency of occurrence of these increments is about the same. You can visually see that there are a small number of increments of very large magnitude (movement) in the market chart. So, that's how it is :)

 
Maxim Dmitrievsky:

Already wrote in the MO topic that this is supposedly done in one go using the inverse Lambert transform

but there is too complicated matrix for me https://www.hindawi.com/journals/tswj/2015/909231/

Although there are packages for R and Py

There is a package in R (LambertW) and it "gaussianises" perfectly. Below are the charts of EURUSD/M20 raw and "gaussianised" logreturn.

require(LambertW)
set.seed(12358)
y1 <- diff(log(pr$close), 3)*100
out <- Gaussianize(y1, return.tau.mat = TRUE)
x1 <- get_input(y1, c(out$tau.mat[, 1]))  # same as out$input
test_normality(tail(y1, 4000)) 
test_normality(tail(x1, 4000)) # Gaussianized

kfc

Fig.1 Raw logreturn data

pfgishftschiau

Fig.2 Processed data

 
Vladimir Perervenko:

There is a package in R (LambertW) and it "gaussianises" perfectly. Below are the charts of the EURUSD/M20 logreturn raw and "gaussianised".

Fig.1 Raw logreturn data

Fig.2 Processed data

well, convert back to kotir and then take the fractional returns from the article and you'll get what Alexander is beating on so inconsolably. Idea.

 
Vladimir Perervenko:

There is a package in R (LambertW) and it "gaussianises" perfectly. Below are the charts of the EURUSD/M20 logreturn raw and "gaussianised".

Fig.1 Raw logreturn data

Fig.2 Processed data

Can you explain why to "gauss" quotes, what advantages it should give, etc., and how to deal with new data? How to work with new "incoming" data using this method?

123456789101112
New comment