Activity Spectrum and AFC of MTS using the Moving Average advisor as an example - page 3

New comment
 
DC2008 >> :

If fast, then yes - filter, but not by activity spectrum, but by AFR. In other words, we simply cut off frequencies at which the Expert Advisor makes losses. And at the second step, we cut off frequencies where the drawdown exceeds the specified limit. That's all: the Expert Advisor is unprofitable and becomes profitable. And I don't care which strategy it uses: as long as it uses the right one.

It becomes profitable on history. There is no continuous group of frequencies with good results.

And frequencies, as we know, "float".

And in this situation it is a transfer from one category (profitable) to another (non-profitable).

can be achieved by shifting the test period by one bar (one day).

 
In what environment do you get this data? Or is there a ready-made tool?
 
DC2008 >> :

If fast, then yes - filter, but not by activity spectrum, but by AFR. In other words, we simply cut off frequencies at which the Expert Advisor makes losses. And at the second step, we cut off frequencies where the drawdown exceeds the specified limit. That's all: the Expert Advisor is unprofitable and becomes profitable. And we don't care which strategy it uses, as long as it uses the right one.

How is it better than history matching?

 
begemot61 >> :

>> how is that better than storytelling?

that's the story. it's just very original. have you never been in touch with the perpetual motion machine designers?

 
jartmailru >> :

How about emulating a quantum computer on a PC? :-) >> I want to feel it.

http://jquantum.sourceforge.net/

 
jartmailru писал(а) >>

Became profitable on history. There is no continuous group of frequencies with good results.

And frequencies are known to "float".

IMHO in this situation, shifting from one category (profitable) to another (unprofitable)

may be achieved by shifting the testing period by one bar (one day).

Let's take it one day at a time:

  • Time is excluded from calculations, i.e. analysis is performed in the space of quantum frequencies - frequency on one axis, and amplitude on another. Each quantum occupies only its place defined for it and does not wander on its axis. In this space only amplitudes change.
  • There is no need to look for a continuous group of frequencies at all. Only those frequencies are needed - where "money lies".
  • Stability to profit does not depend on shift of testing period, because quantum frequencies do not depend on time. For example: if the market emitted frequency 35 on March 1, and frequency 480 on March 7, the frequency, emitted by the market will not change and the analysis will start with frequency 480.
 
mpeugep писал(а) >>
And in what environment do you get this data? Or is there a ready-made tool?

All necessary data is produced by the tester and the analysis in Excel. It is possible to automate this, or better yet: have a function built into the MT - spectral analysis of mts.

 
begemot61 писал(а) >>

How is it better than history matching?

If I understand fitting in the same way as you do - picking the parameters of the MTS so that it becomes profitable at some point in history. Then the proposed (quick) method of analysis allows you to tune (tune like a radio receiver) any MTS to the market, i.e. the robot trades only on those frequencies that are profitable. We do not interfere with the design of the MTS (radio receiver) and do not re-solder anything.

===

"Quick and dumb" to turn a losing strategy into a profitable one does not mean to create a profitable trading strategy, it has remained unprofitable as it was. It just stopped losing.

===

The most correct (though it is long and difficult) way is to study the behavior of mts in the zones of maximum activity (interference) and, as a result, to develop a truly profitable strategy.

 
DC2008 >> :

Let's take it one step at a time:

  • Time is excluded from calculations, i.e. the analysis is carried out in the space of quantum frequencies - on one axis the frequency, and on another the amplitude. Each quantum occupies only its place defined for it and does not wander on its axis. In this space only amplitudes change.
  • There is no need to look for a continuous group of frequencies at all. Only those frequencies are needed - where "money lies".
  • Stability to profit does not depend on shift of testing period, because quantum frequencies do not depend on time. For example: if on March 1 the market emitted frequency 35, and on March 7 - frequency 480, the frequency of testing starting not from March 1 but from March 7 will not change and the analysis will start with frequency 480.

I.e. there is a certain MTS containing period dependence. In fact, you are saying that behavior of MTS (trading system) relative to the condition that the period is defined correctly is invariant... So, if I defined a period and put a digit into MTS, then MTS will work *right* and the same way as it did on previous data?

 
Something similar has long been implemented.'Self-learning EXPERT(lsv)'
123456789
New comment