Machine learning in trading: theory, models, practice and algo-trading - page 848

 
Maxim Dmitrievsky:

Returns are easily converted back to price

The problem is that returns with a small lag are stationary but give a forecast for only 1 bar

with larger lags they are non-stationary but give n bars forecasts

if you manage to get a stationary return with a long lag, it's a kind of a grail

To better understand the value of Erlang flows, here is the beginning of the research:

https://www.mql5.com/ru/forum/221552/page270

Uneven tick arrivals, their delay and filtering negatively affect the results of studies and make it difficult to correctly evaluate the data statistically.

There is a possible solution to the problem by applying Erlang flows:

http://stratum.ac.ru/education/textbooks/modelir/lection29.html

The first link points to the problem, the second to a possible solution.

Here,https://www.mql5.com/ru/forum/137046/page3#comment_3471618 the issue of operating time was raised.

You can try to modify the sifting algorithm, moreover it is not always easy to determine the distribution of time intervals between ticks of a particular broker, the main thing is to define the task of getting the stationary data flow and the normal distribution of the transformed price flow.

От теории к практике
От теории к практике
  • 2018.03.27
  • www.mql5.com
Добрый вечер, уважаемые трейдеры! Решил было на какое-то время покинуть форум, и сразу как-то скучно стало:)))) А просто читать, увы - неинтересно...
 
Novaja:

imho, research about the game in ticks from dts is a useless exercise

but about Erlang it may be useful

Alexander has already written all that is necessary. I just haven't done anything yet, since I'm doing my other stuff.

you can sample time intervals from here https://www.mql5.com/ru/docs/standardlibrary/mathematics/stat/geometric

I'll leave it here so I don't forget it.

Документация по MQL5: Стандартная библиотека / Математика / Статистика / Геометрическое распределение
Документация по MQL5: Стандартная библиотека / Математика / Статистика / Геометрическое распределение
  • www.mql5.com
//| Script program start function                                    | //|  Calculate frequencies for data set                              |                               //|  Calculates values for sequence generation                       |
 

It's not appropriate for such a prominent branch to gather dust in a corner. Rise and shine!!!!!

Have you run out of ideas? Grail seekers...

 
I switched to selling indicators in order to save money for a normal deposit...
 
Evgeny Raspaev:
I switched to selling indicators to save some money for my depot...

Wise decision! Need startup capital.....

 
Mihail Marchukajtes:

Wise decision! Need Startup Capital.....

))))) And you scolded me for selling))))

 
Evgeny Raspaev:

))))) And he scolded me for selling))))

I do not remember, maybe it was in some context. Even if it was true, I still think that trading with indicators, which bring profit, is nonsense. In my case I'm planning to hold a seminar (for a fee) where I will talk about my understanding of the market. As it really is. And share the strategy. Also exclusively for the accumulation of seed capital.... With which you can already at least just live ... :-)

 
Vizard_:

What's the big deal.
Sell your homeland)))

I'd hate to, but there's no way out.... :-(

 

Way to go!!! It feels like a breakthrough in the MO. Congratulations.... coule....

 
So no one knows how to determine the importance of predictors in regression? (other than correlation)
Reason: