Machine learning in trading: theory, models, practice and algo-trading - page 2504

 
JeeyCi #:

to the foundation will not work, the foundation rules itself... The foundations of classical economics are models, but on their basis the regulator makes decisions - the one who will not continue the trend if/when from his/her professional point of view it ruins social well-being (+/-) and social-economic development... I'm not getting into this topic, having lived (witnessed in the market) the full economic cycle (and seen the decisions made by the regulator) -- it's not up for discussion (but macroeconomics textbooks have everything) -- we do not measure, they measure and make decisions themselves, the retail trader can only accept the given/facts as it is

Events in real life are the cause of price behavior as a function of causes, a consequence of causes. There are too many Events and their influence on the price is far from linear and often impulsive. The regulatory impact is more or less clear and long, but besides it there are too many Events. And understanding the relationship of events and their impact on the price is a normal task, and understanding it gives more stability to the profit.

A good example, the famine in Europe in 1664-1667 due to summer frosts has long been known, and the cause - a volcanic eruption somewhere in the ocean was found relatively recently. (I can be wrong about the dates)

 
Valeriy Yastremskiy #:

Events in reality cause price behavior as a function of causes, a consequence of causes. And there are too many Events and their impact on the price is far from linear, and often impulsive. Regulatory influence is more or less clear and long, but there are too many events besides it. And understanding the relationship of events and their impact on the price is a normal task, and understanding it gives more stability to the profit.

A good example, the famine in Europe in 1664-1667 due to summer frosts has long been known, and the cause - a volcanic eruption somewhere in the ocean was found relatively recently. (I could be wrong about the dates).

And I have ALWAYS said that we need to investigate the cause of the change The causal model, remember? Expectations-Expectation Fulfillment or not-Result that is expressed in data Smile-(OI,Delta,Volume)-Price-Indicators.


P.S. Instead of dashes should be an arrow in the downward direction because ONLY Causal Models have the direction of the solution, one from the other!

 

Maybe I'm going to say something fucked up again, but when I was going home and thought about what I could please you with and a trivial idea came to my mind ....

According to the idea, the trading robot should be strengthened by using different types of networks. That qualitative gain in productivity of a network which is noticeable even to the eye is a combination of different types of networks, but each should stand on its own place.

And now a thought, for your appreciation:

Create a static polynomial running on signals and an adaptive NS following the polynomial, usually networks without a teacher, and voila. PROFIT!!!!

In fact I feel myself as a BRUSSLEY LEE in trading, when I can barely cope with one instrument making 10 thousand models rather than to lead the infinite number of instruments at once making one model for each.

And here is a story that happened to me. When I was invited to a forex presentation as an introductory briefing, I almost fell off my chair :-) When a granny told me that she analyzed all the instruments in the trading terminal for all major currency pairs, a standard set.

I.e., the amount of work may increase in any case, as well as the direction to be chosen. In order to achieve the desired results, it is necessary to merge the efforts because the amount of work increases and one person cannot cope with it alone, when everyone works for one common result with the necessary condition that everyone must do his part of the work and do it with dignity!

 
Mihail Marchukajtes #:

Maybe I'm going to say something fucked up again, but when I was going home and thought about what I could please you with and a trivial idea came to my mind....

According to the idea, the trading robot should be strengthened by using different types of networks. That qualitative gain in productivity of a network which is visible even to the eye is a combination of different types of networks, but each should stand on its own place.

And now a thought, for your appreciation:

Create a static polynomial running on signals and an adaptive NS following the polynomial, usually networks without a teacher, and voila. PROFIT!!!!

Net, net, neural net, nanoset, spider net, seine, it's all for fishing, not for Forex. Take a fishing rod, catch a crucian carp, and give your brain a rest.
 
What is the best way to divide random numbers within some limits into cohorts?
 
Mihail Marchukajtes #:

And now a thought, for the appreciation:

Create a static polynomial working on signals and an adaptive NS following the work of the polynomial, usually networks without a teacher, and voila. PROFIT!!!!

the thought is the same - like the mate of all your jargon... you should get into YOUR topic so as not to mislead people about your emotional instability regarding your nonsense regarding this topic... well if the thought is yours, then you can swear at yourself -- well you sent yourself -- go build a polynomial without a teacher ((((((( ... you'll be profitable... as all such builders.......... be kind, send yourself and your thoughts silently, so as not to desecrate Russian language with your scribbles...

I get it, you only memorized the name of your method without bothering to figure out "what it is"... now I understand what kind of team you are assembling... It's sad to see where this branch is going.

Readthe math(the vocabulary is probably too much for you).

 
JeeyCi #:

There is a problem that I sometimes propose to local mathematicians to solve. The answer is usually a set of rudeness and swearing. Can you break this sad tradition and answer the essence of the question?

Problem: calculate the ACF of a random walk.

 
Aleksey Nikolayev #:

There is a problem that I sometimes propose to local mathematicians to solve. The answer is usually a set of rudeness and swearing. Can you break this sad tradition and answer the essence of the question?

Problem: calculate ACF of random walk.

Does it make sense or is it a test for professional suitability?
 
mytarmailS #:
Does this make sense, or is it a test of professional suitability?

Mostly just to keep the conversation going)

Proficiency (in the sense of mathematics) is too loud for all of us here) at most some meaningfulness of statements)

 
Aleksey Nikolayev #:

There is a problem that I sometimes propose to local mathematicians to solve. The answer is usually a set of rudeness and swearing. Can you break this sad tradition and answer the essence of the question?

Problem: calculate the ACF of random walk.

Random from random is random.

Random walk from random walk random walk.


There is a direct relationship/conversion between the ACF and the spectrum. Taking spectrum of wandering is not obvious, because 1/(f^2), using increments, spectrum is equalized, from side phase is rotated by 90gr.