Machine learning in trading: theory, models, practice and algo-trading - page 1130

 
Renat Akhtyamov:

That is, if you take a window the size of the entire history, it will be a flat from beginning to end?

why?

 
mytarmailS:

why?

just wondering what's going to happen?
 
Renat Akhtyamov:
Just wondering what's going to happen?

(What about looking at it with your eyes? ))

 
mytarmailS:

(What about looking at it with your eyes? ))

now there is an algorithm ;)

let's see?

 
Renat Akhtyamov:

now there is an algorithm ;)

let's see?

 

You can also try to find a directional movement in terms of the algorithm

we will consider a trend when two parameters are less than two

and a flat, as it was, more than 7

gp <- get_parameters(x[ii])
  
  flat <- gp[1]>=8 && gp[2]>=8
  trend <- gp[1]<2 && gp[2]<2

a flat is marked in blue and a trend in pink

it was 50 points in the sliding window

and now it's 200 points


 
mytarmailS:

You can also try to find a directional movement in terms of the algorithm

we will consider a trend when two parameters are less than two

and a flat, as it was, more than 7

a flat is marked in blue and a trend in pink

it was 50 points in the sliding window

Now it's 200 points.


How can you tell in real time when a pattern has started... it's just pictures on the history

 
Maxim Dmitrievsky:

How do you know in real time when a pattern has started... It's the pictures on the history again.

No way in real time, only at the end of the sliding window... But the point is that nobody even knew post factum how to algorithmize this pattern.

Well well, I had a task to algorithmize a flit for my system, I did it) bravo to me, haha)

Igor Makanu:

Why are the non-shaded areas in the 1st figure and the TF on which you conducted the study?

It's the ones that don't fitmy flat / trend pattern

Those are the areas where x1 and x2 were greater than "1" (trend) and were less than 8 (flat)

Those are sort of in the middle between a flat and a trend

 
mytarmailS:

In the real time there is no way, only at the end of the sliding window... But the point is that no one even post factum did not know how to algorithmize this fly, that's the first thing, and the second - these things can be clustered and maybe get not bad signs

Well well, I had a problem to algorithmize a flat for his system, I did it) bravo to me, haha)

no, i will never believe in such a thing - that it will work

 
Vizard_:

This is not a TF.

That's right, this is not a TF, the examples were made on randomly generated prices for convenience and speed of reproducing examples. You can take any TF at all.

Reason: