From theory to practice - page 349

 
Novaja:

Great to have you here, I was wondering how the research is going with Erlang flows? Can we hope to get a normal distribution?

HelloNovaja.

A controversial, unexplored area...

As long as I can attach 2 files for EURUSD in my opinion for April 25-27.

EURUSD.csv - evenly read data with interval = 2 sec.

EURUSD_Erlang_1.csv - exponential readout with p=0.5(average interval also = 2 sec.).

If you look at the distribution of increments of these series, you can immediately notice that for the simplest flow EURUSD_Erlang_1, the asymmetry is nearly equal to 0, unlike the uniform flow.

I will, in due course, complete the research. Patience, my friends.

Files:
EURUSD_Datas.zip  341 kb
 
Alexander_K2:

Hello, sweetestNovaja.

A controversial, unexplored area...

So far I can attach 2 files for EURUSD in my opinion for April 25-27.

EURUSD.csv - even reading of data with interval = 2 sec.

EURUSD_Erlang_1.csv - exponential readout with p=0.5(average interval also = 2 sec.).

If we look at distributions of these series, it is immediately noticeable that asymmetry almost = 0 for the simple EURUSD_Erlang_1 flow, unlike for the uniform flow.

I will, in due course, complete the research. Patience, my friends.

Very glad to see you too)) Tell me, what is the shift in readout to 2 sec. connected with?

By the way,@Maxim Dmitrievsky did an exponential readout for ticks.

https://www.mql5.com/ru/forum/86386/page875#comment_7299394

 
Novaja:

Very nice to see you too)) Can you tell me, what is the reason for switching to 2 seconds in the readout?

By the way,@Maxim Dmitrievsky did an exponential readout for ticks.

https://www.mql5.com/ru/forum/86386/page875#comment_7299394

This is as an example. According to my data, further increasing reading interval, some junk still sits in the uniform flow, while in Erlang flows Laplace(???) distribution appears, i.e. in Excel kurtosis=3, asymmetry=0. A normal distribution in Excel should give kurtosis=0, shouldn't it? I don't see that yet.

Yes, I have. Max is good.

 
Alexander_K2:

This is as an example. According to my data, further increasing the reading interval, there is still some junk sitting in the uniform flow, and the Erlang flows draw a Laplace(???) distribution, i.e. in Excel kurtosis=3, asymmetry=0. A normal distribution in Excel should give kurtosis=0, shouldn't it? I don't see that yet.

Yes, I have. Max is good.

By the way, I've already been asked about it, foreigners are interested..:

What is the distribution of price?

 
Renat Akhtyamov:

By the way, I've already been asked about this, foreigners are interested...:

Laplace distribtion of prices?

I suspect it's not even a Laplace distribution, but a hyperbolic distributionhttps://en.wikipedia.org/wiki/Generalised_hyperbolic_distribution

Why is it so important to achieve a stable distribution of increments?

Clear as day - then all known mathematical power for Gaussian processes, Levy processes etc. - are at your service.

Until such a known distribution of gradients is obtained (and it's impossible to achieve it with uniform readout IMHO), any Golden Grail is out of the question.

There will be a wooden Grail (on the basis of Shelepin's theory, with 1 deal a week, like mine) and nothing more. But, we literally want to rip cash from the tree of life every second, don't we?

 
Alexander_K2:

This is as an example. According to my data, further increasing the reading interval, there is still some junk sitting in the uniform flow, and the Erlang flows draw a Laplace(???) distribution, i.e. in Excel kurtosis=3, asymmetry=0. A normal distribution in Excel should give kurtosis=0, shouldn't it? I don't see that yet.

Yes, I have. Max is good.

Yes, you're right, skewness=0 is good, kurtosis=3 is bad. These indicators refer to Laplace distribution (bilateral exponential).

The only data values (observed or observed) that contribute to the kurtosis in any meaningful way are those outside the peak area; i.e. outliers. Hence, the kurtosis measures only outliers; it knows nothing about "peak".

A distribution with apositive excess of kurtosis is calledleptokurtic or leptokurtic. "Lepto" means "thin."[9] In terms of shape, a leptokurtic distribution hasdensertails.Examples of leptokurtic distributions include inStudent's distribution,Rayleigh distribution,Laplace distribution,exponential distribution,Poisson distribution andlogistic distribution. Such distributions are sometimes calledsuper-Gaussian.

One more distribution you should consider: the hyperbolic distribution.

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

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

 
Alexander_K2:

I suspect not even a Laplace distribution, but a hyperbolic distributionhttps://en.wikipedia.org/wiki/Generalised_hyperbolic_distribution

Why is it so important to achieve a stable distribution of increments?

Clear as day - then all the known mathematical power for Gaussian processes, Levy processes, etc. - are at your service.

Until such a known distribution of gradients is obtained (and it's impossible to achieve it with uniform readout IMHO), any Golden Grail is out of the question.

There will be a wooden Grail (on the basis of Shelepin's theory, with 1 deal a week, like mine) and nothing more. But, we literally want to rip cash from the tree of life every second, don't we?

Well thoughts converge)))) And generalised hyperbolic, the whole problem is in the 'heavy tails', we need to lighten them up, get the normal.

Alexander, if it's not difficult, you already have everything calculated on file, can you drop me a message?

 
Alexander_K2:

I suspect not even a Laplace distribution, but a hyperbolic distributionhttps://en.wikipedia.org/wiki/Generalised_hyperbolic_distribution

Why is it so important to achieve a stable distribution of increments?

Clear as day - then all the known mathematical power for Gaussian processes, Levy processes, etc. - are at your service.

Until such a known distribution of gradients is obtained (and it's impossible to achieve it with uniform readout IMHO), any Golden Grail is out of the question.

There will be a wooden Grail (on the basis of Shelepin's theory, with 1 deal a week, like mine) and nothing more. But, we literally want to rip cash from the tree of life every second, don't we?

I hope you've seen my post justifying the absurdity of HFT forex trading?
 
Renat Akhtyamov:
I hope I have seen my post, that substantiates the absurdity of HFT trading in forex?

I made it in time. Notably, if you try your model closer to midnight, the results will be several times more profitable. It would seem that you could make money, but unexpectedly, the spread will also be larger at that time.

As soon as we have an opportunity to profit, the spread increases. What an unpleasant coincidence.

 
the idea of coming up with some kind of distribution of time intervals, superimposing it on real ticks and getting a Laplacian distribution?

Of course, every idea has a right to exist, but I don't think it's going to work here because it's all artificial.

I don't think that if you impose artificial time intervals on a process without memory, you can make it a process with memory.