Discussion of article "Optimal approach to the development and analysis of trading systems" - page 7

New comment
 
Denis Kirichenko:

You should ask your superiors. Or you'll end up like Sergei Bodrov. I mean banned.

Come on, nothing will happen to me ) the main thing is not to refer to the source in the article ). The main thing for me is the opinion of people ). You can even give me examples of other robots, I will look at them and if I understand the logic, I can write an article and everyone is good in the end ).

 

Respect and thanks to the author. I read it with benefit and pleasure. Don't be picky about the code - it is an illustration of some idea.

I am left unclear with which shoulder you tested your impressive and mysterious code? That's a great result. I hope there will be more details in the future.

 
AMK_robot:

Respect and thanks to the author. I read it with benefit and pleasure. Don't be picky about the code - it's an illustration of some idea...

It's funny to read such a thing on a programmer's forum...

 
AMK_robot:

Respect and thanks to the author. I read it with benefit and pleasure. Don't be picky about the code - it's an illustration of some idea.

I am left unclear with which shoulder you tested your impressive and mysterious code? That's a great result. I hope there will be more details in the future.

There is actually nothing impressive and mysterious ))) the code in general is clumsy ) actually forum members trolled it ) I don't deny it. The EA idea is not working either. I had tonnes of them. It is purely for illustration, only as an example. There will be articles I will show how to write a multicurrency close to the table at the beginning of the article. Forget about this Expert Advisor better. Closer to the New Year there will be a more specific article.

 
Denis Kirichenko:

It's funny to read this on a programming forum....

I was more surprised by "excellent result" ))). The code ***s him DDD. Well, it happens ). What's a shame few people understood what the article was about at all. Maybe I believed too much in the interest of the audience

 
Great article, thank you for spending time publishing this. I have managed to grasp the main idea of system development steps and apply these ideas to my price-action based system. But few parts I didn't understand like Modes idea in general, it will be nice if you explain more in your next articles. 👍
 
Martian4x:
Great article, thank you for spending time publishing this. I have managed to grasp the main idea of system development steps and apply these ideas to my price-action based system. But few parts I didn't understand like Modes idea in general, it will be nice if you explain more in your next articles. 👍
The idea of modes is simply not based on giving more variation to the algorithm. This does not guarantee a result, but it will provide a greater chance of success. here the thing is that there are no guarantees, the maximum that we can squeeze the maximum out of our idea, while you need to remember that if there is no light at the end, then most likely it is worth switching to another idea. I've already got a lot of my article done. Many have not been translated yet, follow the forum news, sooner or later they will be translated

.

 
Theoretically everything is correct. Only in practice it turns out that theoretically correct systems give almost no result. This is not my point of view. These conclusions come from analysing trading methods of successful traders with long-play strategies. My signals are simply divided by the degree of risk. The greater the risk, the greater the probability of drawdown, but the greater the income. ....
 
  • LinearFactor = MaxDeviation/EndBalance
  • MaxDeviaton = Max(MathAbs(Balance[i]-AverageLine))
  • AverageLine=StartBalance+K*i
  • K=(EndBalance-StartBalance)/n
  • n - number of deals in the test

Thanks to the author. I, like many people, have always paid most attention to the linearity of profit growth. I use it for optimisation in this form, though without additional classes and other complications (I don't know how to do that). I calculate the balance and fill the array in OnTradeTransaction() after each trade DEAL_ENTRY_OUT, and in OnTester() I compare it with the ideal line. Works for 1 run. I added the possibility to switch the custom criterion. I also use the variant of mean deviation from the ideal balance line and the variant of ecuaiti deviation from the ideal balance line. I agree that"flatness of the curve" ))) gives the best idea of the system performance in the future compared to other criteria, but of course not 100%.

Well, in OnTester() I also discard unnecessary variants such as "few trades" and "small expectation" by zeroing LinearFactor, so the optimiser does not use such genes either and focuses on "more correct" results.

Thanks again.

1234567
New comment