Results of Strategy Tester does not match with single result (after double clicking entry)

[Marcel Fitzner]

Hey guys,

I thought I give it a try by spending a week on a trading idea that I found interesting to implement in a rather simple indicator
and coding some fancy EA to execute the trade signals.

The results look somewhat interesting, but before I'd like to look into optimizing
I was surprised to see that the backtest results of a single entry do not match with the values shown in optimization results table.

I Double clicked the third entry with a profit of 3378.59 - but as you can see the graph comes nowhere close to that profit.

The backtest tab also shows that the number of trades (32) does not match the Total trades in the table (42).

All my indicators that I'm using in the EA do have the 

#property tester_everytick_calculate

The settings of the optimization run look as follows:

I also ran the whole simulation without "profit in pips for faster calculation" and it had the same issues.

Do you have any idea what I could have missed? Why is that discrepancy so huge? Has the strategy tester become so unreliable?

2-3 years ago I used it extensively but never had such issues before.

Kind regards
Marcel

[Stanislav Korotky]

Most likely, you have a bug in your EA. Please, check all possible initializations: variables, arrays, global flags (_LastError resetting, for example), etc (including libraries if you use some). Something is not initialized properly.

[Alain Verleyen]

The Strategy Tester is very reliable.

Such issue has been discussed numerous times, most of the time it's impossible to say anything useful as we don't have a clue about the code used.

When it was possible to investigate, it was always a coding issue.

I also had a case where the problem was a bug in the parameters transmission using "Single Run" (some parameters were not the same). So you should check that firstly.

[Marcel Fitzner]

Hey Alain ! :)

Yeah, I would think so, too (that the ST is very reliable). It seems I got a little rusty and will have to dive deeper into my code then.

Thank you for the hint with the "single run". I'll keep my eyes open wide.

[Marcel Fitzner]

I reduced the code base substantially and simplified wherever I could. Also, 90% of the entries in the optimization table seem to be OK.

However, the entries with the best profits are completely bonkers. No clue why.

Does anyone want to spend a paid hour or two and check the code? I share the whole code base - it's not too huge anyways and the indicator (MACD Impulse) is pretty straightforward.

If this is not the right place to ask for a direct contact here,
I will adhere to the freelancer market although I usually don't like to work with random guys.

At least on Upwork I seem to have a special luck with many applicants who pretend they can code and know how to fix issues, but in fact just write down hours and finally give up...

[Alain Verleyen]

Marcel Fitzner #:

I reduced the code base substantially and simplified wherever I could. Also, 90% of the entries in the optimization table seem to be OK.

However, the entries with the best profits are completely bonkers. No clue why.

Does anyone want to spend a paid hour or two and check the code? I share the whole code base - it's not too huge anyways and the indicator (MACD Impulse) is pretty straightforward.

If this is not the right place to ask for a direct contact here,
I will adhere to the freelancer market although I usually don't like to work with random guys.

At least on Upwork I seem to have a special luck with many applicants who pretend they can code and know how to fix issues, but in fact just write down hours and finally give up...

If you provide me all what is needed to reproduce the issue, I can check to confirm it's related to the code or not, free of charge of course.

[Marcel Fitzner]

Thank you for that kind offer, Alain. I'm preparing a Github repo at the moment. Will let you know the details soon.

[Marcel Fitzner]

https://github.com/CreativeWarlock/MACD-Impulse

Let me know if the EA won't compile due to references leading to missing files. I will add them to the repo asap.

Please send me your public wishlist or a paypal link so I can thank you formally for your efforts!

If you have any questions about the code (because some parameters don't make much sense to you), we can make a brief meeting and I can explain it in short (rather than lengthy posts of questions and answers here).

However, if you found bugs or want to contribute a change to my EA, then I would like it if you fork my project into a branch and
make a Pull request to main, so we can easily discuss relevant lines of codes (also code smells).

(If the Github repo doesn't allow you to make a contribution, I might need you to make a contributor, in that case DM me and let me know your email addy.)

Additional notes:

I added:

- 'Profiles/Tester' with the *.ini files for the EA, so you can start right away with my latest optimization settings.
-  'MetaQuotes-Tester-Logs' with the latest log file with one of the supposedly best profit results (3.5k) where in face the single run yielded only something like 1.1k