From theory to practice - page 1499

 
Vizard_:


Welcome back, Maestro!

Yes, I've seen that it's possible - you've shown it to me (and not only me) before. But, by golly, I don't understand how to do it yet...

 
As an approximation, a similar picture is obtained by dividing the price by the volume. Volume per forem can be modelled roughly as say the sum of squares of x-l minutes over a period
 
vladevgeniy:
As an approximation, similar images are obtained by dividing the price by the volume. The volume per forem can be modeled roughly by say the sum of the squares of x-l minutes over a period

Can you demonstrate this with an example? Please, if it's not too difficult, of course.

And pulling a profit out of a stationary series like Warlock's I will show you how to do it.

 
Alexander_K:
Lousy pound is literally wrecking my TS... It's a shame...

Only trade crosses. There are fewer trends on them than on the majors.

 

I don't have one ready right now. I am too lazy to write the code. And I personally did not get the point, it is important that it still somehow has the growth of deviation, either linear or exponential, or whatever. But here we have a kind of almost stationary series, but it's as if it is always stationary)) I could not find any points.

 
vladevgeniy:

I don't have one ready right now. I am too lazy to write the code. And I personally did not get the point, it is important that it still somehow has the growth of deviation, either linear or exponential, or whatever. But here we have a kind of almost stationary series, but it's as if it is always stationary)) I could not find any points.

Try to write the formula, if it's not difficult.
 
vladevgeniy:

I don't have one ready right now. I am too lazy to write the code. And I personally did not get the point, it is important that it still somehow has the growth of deviation, either linear or exponential, or whatever. But here I got some kind of stationary series, but it's kind of always stationary.

Okay. It's just the way I am - if you have a desire.

The main idea of Koldun (actually, as well as I had at the initial stage of this thread) - to transform the original series of increments to a stationary form. When the probability distribution is symmetric and has a constant variance.

In this case, indeed, the process has no drift and the profit is easily extracted using the cumulative sum of the increments.

But, how to make such a conversion! Do I know?!!! I have no idea.

 
Roman Kutemov:
Try to write the formula

Oh, man... I've already written it down)) I'm just giving you an example of how to emulate volume, it's a matter of opinion. Here we go...

It's a price increment indicator, like a zigzag up and down. Without volume it looks so-so.


And this one has only division by volume accumulated during the period. It looks more stationary)))

The formula (H-L)/(V*K); well, if you want to know in general) IMHO it was clear anyway

 
vladevgeniy:

Oh, man... I've already written it down)) I'm just giving you an example of how to emulate volume, it's a matter of opinion. Here we go...

It's a price increment indicator, like a zigzag up and down. Without volume it looks so-so.


And this one has only division by volume accumulated during the period. It looks more stationary)))

Looks a little bit. Now take the cumulative amount over some period of time, calculate the standard deviation using the formula =sqrt(D*t), multiply by some quantile of the Gaussian distribution. You will get to a stationary channel relative to 0. When crossing the upper limit - SELL, when crossing the lower one - BUY. Exit from the trade - when returning to 0. That's all.

 
Alexander_K:

Welcome back, Maestro!

Yes, I've seen that it's possible - you've already shown it to me (and not only me). But, by golly, I don't understand how it's done...


I missed it again. Let me see what it is.


Alexander_K:

Looks like a bit. Now take the cumulative sum over some period of time, calculate the standard deviation using the formula =sqrt(D*t), multiply by some quantile of the Gaussian distribution. You will get to a stationary channel relative to 0. When crossing the upper limit - SELL, when crossing the lower one - BUY. Exit from the trade - when returning to 0. That's all.


Drawing thousands of nice incremental sums with a long interval without any quantiles. The problem is the same, not always the price goes up when going over the lower boundary.