How do I identify the patterns on which the ready-made TS shows a profit? - page 6

New comment
 
zaskok3:
I wrote a random TS and it shows a profit. But I can't understand what patterns it is based on. I need this in order to create a conscious TS based on these regularities and, consequently, to have adequate ideas about refinement.

An accidental TS is when you feel it without understanding what you are doing. For example, let me insert the input parameter here and prooptimize it. And lo and behold, the picture will be completely different. But I cannot understand what this parameter exploits.
I have read all 5 pages and did not understand - had the author understood what he was doing and knew on what laws the TS is based, or no? If yes, then his unknown law is the result of"and let me add the input parameter here".
 
lilita bogachkova:
I have read all 5 pages and did not understand, did the author understand what he was doing and knew on what regularities TC is based, or not? If yes, then his unknown regularity is the result of"and let me insert an input parameterhere".
:)
 
zaskok3:
Statistical studies on the level of dependence of parameters can of course be done in the same R. It's not about the tools, it's about the approaches to explaining the workability of actually randomizing one's TS.
you stress "your own" so insistently that I'm starting to have doubts.... yesterday "my" last week " can't figure out how it trades" ... but in any case it's more complicated than the black box, at least you know what's in it, and here you don't even know it

Sometimes people find treasure, you can try to understand what led to the discovery, in order to have a stable income from the treasure, but something tells me that it would be better to adequately sell the find and use it as seed money for a less risky business than forex
 

I wonder - was the author aware of 'debugging' in the metaeditor when he wrote the first post?

It is quite difficult to avoid mistakes when writing relatively complex programs. MetaEditor has a built-in debugger that helps to solve this problem.

Thedebugging of applications consists in the possibility of the step-by-step program execution, calculating the values of local variables and setting breakpoints in a given position.

 
zaskok3:
Wrote a random TS, it shows profit. But I can't figure out what patterns it is based on. I need it to create a conscious TS based on these regularities and therefore have an adequate idea of how to fine-tune it.
"I ran a random code I found on the net, I want to understand its logic".

To understand it you need to do manual trading instead of programming

Know what and why the price and indicators show, not only their formulas and ways to call them, although even then there is no 100% chance to repeat the system, but I have met people who give very good techniques based on trading history
 
David Azizian:
An EA can have many conditions for entry, but there is an overriding condition. If it is removed the order will not open.
This is where I am wrong. An order may open when any of the conditions are met if the other conditions are removed. It will simply open in the wrong place. So the name of the pattern must include all the conditions. There may be unnecessary useless conditions. I think we should drop them out of the EA one by one and see what happens.
 
The thread has taken on a categorising characteristic. You can already get an idea of the level of perception of algotrading by most of the people who have spoken. I am glad that we have found common ground in our understanding of the problem.

Thanks for the discussion.
 

You have been given an EA with the source code. The TS shows a profit. You look into the source code - it is very difficult. You realise that it will be extremely difficult to understand the logic to get to the point of explaining it to yourself that you can repeat it from scratch.

Will you bet on the real thing?

And if you do manage to understand the algorithm, but it seems to you, to put it mildly, strange: you do not understand how and what the author thought he was smoking. But you have understood the algorithm and have been able to repeat it, managing to prepare the combat unit for all sorts of contingencies.

Are you going to bet on the real thing?

And if you have machine learning, by feeding a bunch of parameters to the input and using different mathematical models of learning (NS and other). We obtained the source code with a bunch of coefficients. At the same time TS shows profit on history and forwards. The algorithm is clear, just a huge set of weight coefficients. So, there is practically no human logic, but there are many-many numbers that give positive results in the tester.

Will you bet for real?

 
zaskok3:

You have been given an EA with the source code. The TS shows a profit. You look into the source code - it is very difficult. You realise that it will be extremely difficult to understand the logic to get to the point of explaining it to yourself that you can repeat it from scratch.

Will you bet on the real thing?

And if you do manage to understand the algorithm, but it seems strange to you, to put it mildly: you do not understand how and what the author thought he was smoking. But you have understood the algorithm and were able to repeat it, managing to prepare the combat unit for all sorts of contingencies.

Are you going to bet on the real thing?

And if you have machine learning, by feeding a bunch of parameters to the input and using different mathematical models of learning (NS and other). We obtained the source code with a bunch of coefficients. At the same time TS shows profit on history and forwards. The algorithm is clear, just a huge set of weight coefficients. I.e. there is practically no human logic, but there are many-many numbers that give positive results in the tester.

Will you bet on real?

  1. Yes
  2. Yes
  3. I don't think so.

Although it would be more correct to allocate the risk capital:

  1. 50%
  2. 30%
  3. 20%


But this is very tentative. Much depends on the quality of the forwards, the tightness of the parameters, the similarity of results on other instruments, etc.

 

zaskok3:
Интересно, что большинство ТС в мире, принесших прибыль, написаны именно по этому принципу - случайная ТС. 

I've only done it once. But there are more than one "hmm, but if you do it that way, it should work" algorithm.

zaskok3:

Will you bet on the real?

Dvvlvdno? Give me a really profitable black box with the possibility of tests, I'll find what to do with it.

Regarding the starting theme. You have to understand that there is nothing but common phrases you will not throw here, and will not be able to, not on what. You have an algorithm, an exploitable pattern inside. All that remains is to cut it out. It is easier to cut it off.

1234567
New comment