The market is a controlled dynamic system. - page 7

 
avtomat:


No. No, this is just an example...

I haven't solved the problem yet ;)

The marked area is the realignment area. Having passed it, move on to the current scenario, etc.

.

in a slightly different representation


ok ok... but only 1 BP will be used (one valued pair for example)? if so, what is the advantage of this approach over the same neural network? (of course we can retrain it on every bar)...
 
Vizard: ok ok... but only 1 BP will be used (one valued pair for example)? if so, what is the advantage of this approach over the same neural network? (of course we can retrain it on every bar)...
The advantage may be in the transparency of the model. I would give a lot just for this transparency and logical consistency.
 
Vizard:

ok ok...but only 1 BP will be used ( one shaft pair for example )? if yes, what is the advantage of this approach compared to the same neural network ? (of course we can retrain it on every bar)...

Mathemat The advantage may be in the transparency of the model. I would give a lot just for that transparency and logical consistency.

exactly right,

And neural networks don't allow to deliberately change object properties, such as the degree of oscillation or the order of statism, etc., and as a consequence many research tools cannot be used.

.

zy

but it should be said that neural networks at first interested me as something new (it was long ago), but rather quickly disappointed... although i am not familiar with innovations in this field, maybe i missed something...

 
yosuf:
It always seems to you that I'm not in the loop, although we are discussing dynamic systems, which undoubtedly, in my opinion, include the market. It is necessary to have time to produce controlling actions while the Expert Advisor is running, in real time; otherwise we will again be managing historical data. The author says that I also need to think about implementing a management unit, for which I must follow the scheme described by the author on p. 3 and manage to consider several options simultaneously, develop Actions for one of the selected directions and continuously optimize the Expert Advisor by only one criterion - maximizing the profit, though Vinin says the corresponding resources must be available for this and does not exclude the possibility in principle of doing it. Apparently, this is how the principle of controlling impact should be, although maybe the author would like to explain his point of view on this issue and clarify the situation.

Yusuf, of course I don't have a ready-made solution for your case, but considering the properties of your indicator (as far as I can imagine), on the basis of which your Expert Advisor is built, you can and probably should consider and test a voting scheme for several TFs or several windows on one TF. For example, like this:

.

 
sever30:

what the hell are those who don't understand the formula to do?

what is the topic of discussion, dear Automat?


smoke bamboo. (of course it is manageable if every bazaar is regulated and brigadiered and billions are left to the chaos)
 
Yeah, yeah. Into the abyss of chaos... that's for sure ;)
 
avtomat:

absolutely right,

and neural networks don't allow us to purposefully change object properties........


I agree about transparency....but I meant efficiency... Any controlling factor we will look for and calculate on the previous history (by sliding window over n bars for example)... then feed it into the model ... we can do the same with the grid and any other algorithm ... i.e. if predicting MA or price - just input a line with already taken into account - the same degree of oscillation (it may also be dynamic and calculated by a sliding window for example) or .... the grid or whatever it chooses in the end ... that's what I mean ... the problem is as usual - to find these controlling factors ... that will really work ... and where to put them I think is not so important ...

 

You are talking about forward-looking prediction. The task is: "Based on the Y-cause, determine the Z-effect".

 

I put the question differently: "Determine the X-cause leading to the Y-effect"

 
avtomat:

I put the question differently: "Determine the X-cause leading to the Y-effect"


Well, we can see it in regression analysis, for example, as a regression coefficient...or something else (we get the answer as a function, for example)...that is, as a teacher we put in what we want to end up working on, and as an input - what we look for dependence on...then we look...