Optimise an EA and get the best of the optimised ones. - page 43

 
Aleksey Vyazmikin:

Can you be more specific about what the problems are? Everything seems to be working for me, so I wonder if there's a hidden bug that you've run into.

I don't have any bug! All works fine.

I simply put a breakpoint in OnTesterPass() function and it does not stop there. Although everything is working normally, the function is called and all statements are executed.

The debugging just gets very annoying, I have to write TRACEs and look through the log-file to see how and what is going on.

But, in general, all difficulties have been solved, I don't have time today, but I'll get to the "finish line" tomorrow.

Anatoly, I've studied all these articles carefully, and I'm doing everything as suggested there. Everything's fine, everything works, I'll send you complete stats file for each pass. I just don't understand why I can't get into OnTesterPass() function in debugger.

 
Georgiy Merts:

There is no bug ! Everything is working fine.

I just put a breakpoint in OnTesterPass() function and it doesn't stop there. Although, everything works fine, function is called and all statements are executed.

The debugging just gets very annoying, I have to write TRACEs and look through the log-file to see how and what is going on.

But, in general, all difficulties have already been overcome, I don't have time today, but I will reach the "finish line" tomorrow.

Anatoly, I've studied all these articles carefully, and I'm doing everything as suggested there. Everything's fine, everything works, I'll send you complete stats file for each pass. I just don't understand why I'm unable to enter the OnTesterPass() function in the debugger.

I understand that OnTesterPass() is run from a copy of the Expert Advisor that is additionally created when running optimization and I cannot enter this copy from the original, while debugging is performed on the original.

 

I wonder how long it will take you to realise that this approach doesn't work?

You think it's that simple? Yes, it is simple. But it's not.

 
TheXpert:

I wonder how long it will take you to realise that this approach doesn't work?

I mean, how is it "not working" when there are absolutely concrete results ?!!!

Look. (Let's keep it on a first name basis, we've known each other in absentia for a long time).

Until six months ago, I had three levels of problems.

  1. What would I come up with to make the TS work ? I read about the "beautiful TS", start looking at the chart, and I can already see by eye that it's fucking working. Bummer. Sometimes "something works". I make an Expert Advisor and see that the TS does not work. There are two ways forward - first, go to the beginning of the cycle, and look for something else. The second way - start "adding crutches" - filters, conditions, limitations. As a result, we finally get something working in the tester - and then we reach the second level of problems.
  2. It is nice in the tester. But will it work on a demo account? When we set it on a demo account, the TS usually "goes down" right away. Then we can go back to the beginning of the cycle, or we can start adding "crutches". And if on the first level - everything depended on the speed of writing the code and the speed of tester, then on the second level the time is infinitely stretched. In most cases - you get convinced that the TC does not work. In those few cases where the TC works - you get to the third level of problems.
  3. We have positive results on the demo-account ! But will it work on a real account? We bet on the real account, and TS usually starts to fail. Again, it is a bummer. And only in a small number of cases "something comes out". I must say that I never reached this third level six months ago. But with TC League I not only reached this third level, but I'm even approaching the state of "something goes".

With the TC League - I solved two levels of problems at once ! In fact, half !

What's more, also solved the diversification problem on the third level.

And you call that "the approach doesn't work" ???

Even half a year ago I almost never rose above the first level, and all the time I was thinking - "well, what can I do to make the TS show good results at least in the tester" (even in my tester it was very bad - either the TS was losing, or it had dozens of crutches and settings, which made it terribly unstable). Now the question "what to invent" simply does not exist at all. Everything has already been invented! Moreover, it works not only in the tester, but on a demo account as well!

With League - I've stopped looking at the chart altogether. All the attention is only on the reports and over-optimisation. There's never a question of "what to come up with" - there's just shit to do. There's no way I can call it "the approach doesn't work". I think it's working very well.

I take it that by "approach works" you mean a one-button "chop dough" expert. Alas, the TC League is not a silver bullet. It is a solution to the first two levels of problems. The last level is the one that remains. But, it remains in any case, for any approach - so there is no difference here.

 
I'm not insisting, just hinting unequivocally. After all, it's your time.
 
TheXpert:
I'm not insisting, just hinting unequivocally. after all, it's your time.

Stop "yammering".

And you'd better explain what "the approach doesn't work".

 

I've put in

8GBPUSDChnTrendSARNot allowed SL
 

The latest (so far) TC on over-optimisation will be

EURCHF EMATrendSP

The rest are working, thanks to the man from another forum (well, there are the following symbols, but they will be tested after I finish the statistics, starting tomorrow)
 
Georgiy Merts:

The last (so far) TC on over-optimisation will be

EURCHF EMATrendSP

The rest are working, thanks to the man from another forum (well, there are the following symbols, but they will be tested after I finish the statistics, starting tomorrow)

So should I cancel the optimisation?

 
Aleksey Vyazmikin:

So should I cancel the optimisation?

No, no. It's fine. If you put the file out, the regcode will be added.

(I'm working a little "ahead", so there's no "overlap" yet).

Just dealing more with the code for now. Basically, it's almost done, tomorrow I'll post you the experts who know how to write the statistics file.