I will write a bespoke expert - page 5

New comment
 
As described by aftoroffs - not working)))
 
Lemyx писал (а) >>

What do you call a "profitable pattern"? And how do you distinguish it from an unprofitable pattern?

Well, there's the notion of a diversion - for example, when the price makes a new high, but the indicator doesn't. But the history shows a lot of such patterns, both good and bad. How to divide?

just like with patterns:
let's say a methodology has been devised to digitise a divergence. Each distinctive digitization is a pattern.
then we write a script/advisor and run searches for each digitization.
The usual way is to convert the patterns into a linguistic series - i.e. assign a literal to the pattern and write it to a string.
It turns out to be of the eeeeeddddddddddddddddddddd type where "eee" is empty for the given method, d, D are different types of divergence
We run the debugging Expert Advisor to enumerate profitable combinations of patterns, for example "ddeeD" and "dedeeD" are profitable, etc.

 
Korey писал (а) >>
I do not work as described by aftoroffs :)))

How do you work? On patterns? Just on patterns or on diversions? Or maybe with some tricky filter? By hand or by mts?

And is it true that profitable patterns from history often trigger future profits, has anyone checked? I tried to implement a method of such trajectories, but was disappointed in the idea even before I started.

In fact I spent half of my life with diverters, they do not give me much confidence :(

 
Korey писал (а) >>

just like with patterns:
Let's say a method of digitising a divergence is invented. Each distinctive digitization is a pattern.
Then we write a script/advisor and run searches for each digitization.
The usual way is to convert the patterns into a linguistic series - i.e. assign a literal to the pattern and write it to a string.
It turns out to be of the eeeeeddddddddddddddddddddd type where "eee" is empty for the given method, d, D are different types of divergence
We launch the debugging Expert Advisor to search for profitable combinations of patterns, for example, "ddeeD" and "dedeeD" are profitable, etc.

Hmmm... Pretty interesting thoughts on the subject. That's something to think about and put somewhere :-)

 
Lyfesh писал (а) >>

How do you work? On patterns? Just on patterns or on diversions? Or maybe with some tricky filter? By hand or by mts?

And is it true that profitable patterns from history often trigger future profits, has anyone checked? I tried to implement a method of such trajectories, but was disappointed in the idea even before I started.

In fact I spent half of my life with diverters, they do not give me much confidence :(

100% - Not true, there is only probability.
The probability is outweighed by a steady deposit, and ... well in general der Ordnung der Ordnung und noch ein mal Ordnung.

 
Korey писал (а) >>
As described by aftoroffs - not working)))

What do you mean by that? Is there some kind of pattern-based system?

 

A question to those who have a lot of experience in writing EAs. Has anyone tried to implement an EA based on MGUA (Group Method of Argument Counting)?

 
Lyfesh писал (а) >>

A question to those who have a lot of experience in writing EAs. Has anyone tried to implement Expert Advisor based on MGUA (Method of Group Accounting of Arguments)?

More details on this point.

 
Lyfesh писал (а) >>

A question to those who have a lot of experience in writing EAs. Has anyone tried to implement an EA based on MGUA (Group Method of Argument Counting)?

There is a desire to join the 'How to make an automatic multiple test?:) on picking weights in the tester. :)

And simpler still, a 'black box' generating dll and a binder from klot.

But in this place I would like to explain in more detail.

"Xforex formula. ;) After training, with given inputs, something like:

/* inarray[5] is iMomentum */
/* inarray[6] is iADX_MODE_MAIN */
/* inarray[7] is iADX_MODE_PLUSDI */
/* inarray[8] is iADX_MODE_MINUSDI */
/* inarray[9] is iAC */
. . . 

X6 =  2 * (inarray[5] - 98.79059) / 2.203308 -1;
X7 =  2 * (inarray[6] - 6.920272) / 82.10947 -1;
X8 =  2 * (inarray[7] - 0.3410268) / 43.67283 -1;
X9 =  2 * (inarray[8] - 1.141973) / 40.24021 -1;
X10 =  2 * (inarray[9] + 4.32547E-03) / 7.31606E-03 -1;
...

netsum += -1.989235590738936E-002*X27;
netsum += 0.2095882309912516*X53;
netsum += -0.139030684215685*pow(X1,2);
netsum += 43.53590893128754*pow(X15,2);
netsum += 53.92228444532424*pow(X19,2);
netsum += 0.2041931003133029*pow(X1,3);

But:

Unfortunately, it is almost impossible to find a specific answer in the literature on how to apply the MSUA successfully. The reason for this is the fact that the MSUA uses many different parameters, and there is no "standard" set of these parameters.

However, the data is out of date :(.

 
SergNF писал (а) >>

Feel like joining the 'How to make an automatic multiple test?:) on picking up the scales in the tester. :)

And simpler still, a 'black box' generating dll and binder from klot.

"Xforex formula. ;) After training, with given inputs, something like that:

But:

But the data is old :(

And this is the same or not.

http://library.graphicon.ru/pubbin/view_prop.pl?prop_id=19

123456789
New comment