Machine learning in trading: theory, models, practice and algo-trading - page 2753

 
Valeriy Yastremskiy #:

What will fractalism provide?

The properties of fractal series can give insights

Been reading Mandelbrot, generally not bad, you can start to see the "patterns" that are described by this theory

These patterns are different, but share common properties

Predictive power is of course unproven.

 
Maxim Dmitrievsky #:

including, but it is necessary to define their nature and describe them somehow in general in order to use an adequate method including MoD

in my opinion, fractal theory is closer to the body in this case

and it's already fractal, self-similar...

exactly the same graph on the scale of minutes:

and both ways, from any starting point :-))

price tends to minimise deviations (the market, after all, bargaining at all levels and time horizons), so pay attention to where these deviations are maximum

 
Maxim Kuznetsov #:

and it's already fractal, self-similar.....

exactly the same chart on a minute scale:

and in both directions, from any starting point :-)

the price tends to minimise deviations (market, after all, bargaining at all levels and time horizons), so pay attention to where these deviations are maximal.

You know what you're doing, you've even invented an indicator.

I'd be interested in processing this within the framework of the MO. And other properties. Maybe I'll do something later.

OK, I'm off, I'll write something with examples later.
 

Small addition: the SB differs from the classical SB in general by the fact that flatts are forbidden. For economic reasons, a long movement along the horizon is impossible.

 
Maxim Dmitrievsky #:

Properties of fractal series can provide insights into

Been reading Mandelbrot, generally not bad, you can start to see the "patterns" that are described by this theory

These patterns are different but share common properties

predictive power, of course, is unproven.

Properties in our case are very difficult / or even impossible to formalise. Although it is necessary to dig in this direction, there is no answer yet, whether it works or not in our case. For others, including weather, it works.

 
Maxim Kuznetsov #:

points in history beyond the classical SB are worthy of special attention


comet

;)

 

another way of determining the reference points, through polynomials of degree 7-9 (not 2-3, but not excessively, or we will just inadvertently smooth out):

- take a rather large history (many, many bars)

- somehow define or specify the parity of the degree of the polynomial

- assign weights to bars - from the smallest for old bars to the largest for new ones. Linearly or exponentially, as long as they are not equal.

- apply the ISC

voila - we search for valid bars in the vicinity of special points (vertices and inflections).

 
Maxim Kuznetsov #:

another way to determine the reference points, through polynomials of degree 7-9 (not 2-3, but not excessively, or we'll just inadvertently smooth out):

- take a rather large history (many, many bars)

- somehow determine or specify the parity of the degree of the polynomial

- assign weights to bars - from the smallest for old bars to the largest for new ones. Linearly or exponentially, the main thing is that they should not be equal

- push the ISC

voila - we search for valid bars in the vicinity of special points (vertices and inflections)

These repertoires will be bifurcation repertoires, and after them we should look for attractors and adjust the window to them, until the next bifurcation.

in terms of nonlinear dynamics or whatever.

to put this window in the MO and forecast for the near future.

Attractors will be clear only somewhere after the middle of this self-affine piece of the graph, before that nothing will be clear, or a larger or smaller one will act.

there is probably an algorithmically simple way to take everything into account in the training examples, without having to worry about it.

 
Maxim Dmitrievsky #:

These repertoires will be bifurcation repertoires, and after them we should look for attractors and adjust the window to them, until the next bifurcation

in terms of nonlinear dynamics or whatever.

to put this window in the MO and forecast for the near future.

Attractors will be clear only somewhere after the middle of this self-affine piece of the graph, before that nothing will be clear, or a larger one will act.

I don't know why you're so obsessed with windows. About windows - this is to microsoft :-)

You have chosen a point of reference, that's where you should count from. You care about the personal fate of your personal account/series/transaction. That's the main thing.

The longer an account lives, the deeper you have to dig. You yourself, when you were wearing short trousers, thought that a week was close to eternity; now weeks are like a blink of an eye;

 
Maxim Kuznetsov #:

Why are you all so obsessed with windows? The windows thing is for microsoft :-))

You have chosen a point of reference, so you can count from it. You care about the personal fate of your personal account/series/transaction. That's the main thing.

the longer an account lives, the deeper you have to dig. You yourself, when you were wearing short trousers, thought that a week was close to eternity; now weeks are like a blink of an eye;

we need to describe what we want to get out of the window with chips and marks.

it can freeze on these reference points to the next, narrowing-expanding depending on the conditions and with it.

The goal is classification.

I just suggested an alternative to the classical sliding window, since it is so much to everyone's annoyance, especially it can be theoretically justified in some way through the properties of some fractals and some nonlinear dynamics.

It would be an interesting trip to nowhere, but you might get lucky. But since everyone here likes trips, it's as good as any other :D