Discussion of article "A scientific approach to the development of trading algorithms" - page 7

New comment
 
Maxim Romanov:
Then post here the definition of trend from this theory, if it is really more specific, people will be more interested to read your theory.

Definition of trend (as a concept) from the impulse equilibrium theory:

It is a stretch of price movement (a sequence of M-forms) where the main role is played by directional M-forms that have a single (unified) direction.

The main role of these directional M-forms is due to the fact that their total amplitude is predominant in this segment of movement.

 
Aleksandr Masterskikh:

Definition of trend (as a concept) from impulse equilibrium theory:

It is a stretch of price movements (a sequence of M-forms), where the main role is played by directional M-forms, which have one (unified) direction.

The main role of these directional M-forms is due to the fact that their total amplitude is dominant in this segment of movement.

I am quietly writing something like this now. I calculate the difference of extrema taking into account the sign on 132 bars of different TFs and look at the total and average per unit of time. And the average speed and speeds up down. Readable velocities in points per hour are obtained. Interestingly, the average speeds on the lower TFs are higher than on the higher ones. But at the same time their order is close. There is no 10-fold difference.

 

Maxim Greetings. I have read the article and have a question. Do you really consider the charts of equity balances for different instruments attractive?

To me it looks more like a coincidence and when the balance starts to sag you will not have any guarantees that the current losses will be recovered. That is, to make money with this approach you need to know the FUTURE block size, not the current optimal one. Looked through the charts and unfortunately none of the equity could not attract, I say this from the practice of trading. Too big drawdowns, at the moment you won't know if it will recover or not.

1. Think about how to choose the optimal block size for the near future. So that it starts working tomorrow, and not from the week before last. Well, and hold the idea, which has been hovering in my mind for a long time and I even somehow already covered it, but was thrown tomatoes :-)

2. When we perform balance optimisation, the tester selects such parameters that lead to the fastest growth of the deposit balance, but nobody has thought of doing balance optimisation in such a way that the curve would be of this kind.

The task of this optimisation is to find the block size so that the balance curve would be like this, thus we find the point of the beginning of growth, i.e. when the block size is about to bring profit (work) BUT I don't know how to do it in the tester. If it turns out to do something similar I would be grateful to look at it. I have an indicator which has only two parameters and both of them change from 5 to 45 and if there was a possibility to optimise it to this kind of balance curve, I wouldn't even use Skynet. Look, I have marked on the chart these areas that need to be searched, and the main thing is the growth, without dips and drawdowns, unlike your article....

So if you manage to do something in optimisation I will be glad to join this case....

 
Mihail Marchukajtes:

Maxim Greetings. I have read the article and have a question. Do you really find the charts of equity balances for different instruments attractive?

To me it looks more like a coincidence and when the balance starts to sag you will not have any guarantees that the current losses will be restored. That is, to make money with this approach you need to know the FUTURE block size, not the current optimal one. Looked through the charts and unfortunately none of the equity could not attract, I say this from the practice of trading. Too big drawdowns, at the moment of which you won't know if it will recover or not.

1. Think about how to choose the optimal block size for the near future. So that it starts working tomorrow, and not from the week before last. Well and hold the idea, which has been hovering at me for a long time, and I even somehow already covered it, but was pelted with tomatoes :-)

2. When we optimise the balance, the tester selects such parameters that lead to the fastest growth of the deposit balance, but no one has thought of doing balance optimisation in such a way that the curve would be of this kind.

The task of this optimisation is to find the block size so that the balance curve would be like this, thus we find the point of the beginning of growth, i.e. when the block size is about to bring profit (work) BUT I don't know how to do it in the tester. If it turns out to do something similar I would be grateful to look at it. I have an indicator which has only two parameters and both of them change from 5 to 45 and if there was a possibility to optimise it to this kind of balance curve, I wouldn't even use Skynet. Look, I have marked on the chart these areas to be searched for, and the main thing is what kind of growth is going on afterwards, without dips and drawdowns, unlike your article...

So if you manage to optimise something, I will be glad to join this case.....

This is naturally not a combat algorithm, I wanted to show how using a consistent approach to development, you can completely abandon optimisation.
These graphs of profitability without optimisation are made, the profit will be on the instruments, to which there is an active inflow of funds. In its current form, the algorithm is an alternative to strategies of buying attractive assets.
Regarding optimisation, I am not a helper here, because I have given up optimisation. I work in the direction in which the algorithm should trade itself, fully automatically, on any instrument.
The posted algorithm needs to be significantly improved, but I am engaged in another project, and this algorithm is a very small part of a large project, I do not use it in its pure form.
 
Maxim Romanov:
This is naturally not a combat algorithm, I wanted to show you how using a consistent approach to development, you can do away with optimisation altogether.
These graphs of profitability without optimisation are made, the profit will be on instruments that have an active inflow of funds. In its current form, the algorithm is an alternative to strategies for buying attractive assets.
Regarding optimisation, I am not a helper here, because I have given up on optimisation. I am working in the direction in which the algorithm should trade itself, fully automatically, on any instrument.
The posted algorithm needs to be significantly improved, but I am engaged in another project, and this algorithm is a very small part of a large project, I do not use it in its pure form.

I see. It's sad. And for many years no one has been able to optimise the parameters to the reduced form of the balance curve :-(

 
Mihail Marchukajtes:

I see. Sadness. And for years no one has been able to optimise the parameters to the given form of the balance curve :-(

centuries)

 
Valeriy Yastremskiy:

I am quietly writing something like this now. I calculate the difference of extrema taking into account the sign on 132 bars of different TFs and look at the total and average per unit of time. And the average speed and speeds up and down. Readable velocities in points per hour are obtained. Interestingly, the average speeds on the lower TFs are higher than on the higher ones. But at the same time their order is close. There is no 10-fold difference.

In the theory of impulse equilibrium this issue is solved:

- M-form parameters (threshold values, including the speed of price change) have been determined, which allow to predict at an early stage whether the trend will develop or not (of course, probabilistically),

- also regularities in time (and velocity) ratios of M-form parts were revealed, which allowed to determine truly relevant moving averages at any moment of time (on any scale).

 
Aleksandr Masterskikh:

In the theory of impulse equilibrium this question is solved:

- M-shape parameters (threshold values, including the rate of price change) are defined, which allow to predict at an early stage whether the trend will develop or not (of course, probabilistically),

- also regularities in time (and velocity) ratios of M-form parts were revealed, which allowed us to determine truly relevant moving averages at any moment of time (on any scale).

We should look at the theory) for now stupidly algorithm solutions first. And then their variants. If there is a link to the theory, please give it so I don't have to look for it).
 
MetaQuotes:

New article A scientific approach to the development of trading algorithms has been published:

Author: Maxim Romanov

Thanks for the interesting article. There is an custom indicator missing: Max_distribution_v.1.13.mq5 Can you please upload it? Thanks.

 
konorti:
Thanks for the interesting article. There is an custom indicator missing: Max_distribution_v.1.13.mq5 Can you please upload it? Thanks.

Thank you for appreciating the article. Max Distribution V.1.13.mq5 indicator I did not attach it to the article, because it is a unique, author's development and I am not ready to exhibit it for general use. <Deleted>  I warn you right away, for the robot laid out in the article, it is of no value.

123456789
New comment