Errors, bugs, questions - page 489

 
Vigor:
You can see that the Expert Advisor works at the open prices of the hour bar. The last open price was 23-00.

Wrong - m30, you can see it even in my picture. Found a bug on M5.

 
Swan:

The achieved effect of the cutoff is questionable, the number of prints is reduced by 10%, and you have to go to the log for full information...

Let's check, every 10 messages should not be cut off.


Is writing to a file faster than displaying it on the screen?

Definitely faster if we're talking about thousands of lines. The difference in speed is about 100 times and more.
 
Renat:
Let's check, every 10 messages should not be cut off.

Renat, can you make it possible to run the tester to the current time?
It is very useful when the market starts to clearly break down, but we have to wait for the end of the trading day and there is no possibility to re-optimise the system.

Документация по MQL5: Дата и время / TimeCurrent
Документация по MQL5: Дата и время / TimeCurrent
  • www.mql5.com
Дата и время / TimeCurrent - Документация по MQL5
 
crOss:

Renat, can you make it possible to run the tester to the current time?
It is very relevant when the market starts to clearly break down, but we have to wait for the end of the trading day and there is no opportunity to re-optimize the system.

We deliberately made the cut-off date last so that there are no questions of "why two identical passes gave different results". Unfortunately, this is a massive situation that also affects all optimisation results. From the inclusion of the score to the last minute, everyone will get too many problems.

Perhaps we will make an adjustment in the future.

 
Renat:

We deliberately made the cut-off date last so that there are no questions of "why two identical passes gave different results". Unfortunately, this is a massive situation that also affects all optimisation results. From the inclusion of the score to the last minute, everyone gets too many problems.

Perhaps we will make an adjustment in the future.

It would be very nice to get such a setting, although it could be made simpler... At the start of the optimisation the date "current time" is remembered
and until the next optimization starts the tester operates with it as the right boundary of the range, also on test runs. Then the same runs will give the same results. The sheep are well and the wolves are satisfied ))))

What about the last hour dropout on test runs ? Moreover, I noticed that the tester doesn't start from 00:00:00, but when it has to (see picture).

Further, when creating a multi-currency Expert Advisor I faced a problem.
During testing, the quotes obtained for a specific symbol at the same time vary depending on the name of the chart (see picture).
The total system result for many symbols may significantly change due to such price fluctuations, especially for timeframes less than one hour (H1).

 
crOss:

Moreover, I noticed that the tester does not start at 00:00:00 but when it has to (see picture).
The tester starts tests at 100 bars from the beginning to avoid potential errors when working with an empty history. Therefore, the test does not start at 00:00:00
 
crOss:

Further, I faced a problem when creating a multicurrency Expert Advisor.

When tested, the quotes obtained for a particular instrument at the same time change depending on what name the chart has (see the picture).
The total system result for many symbols may significantly change due to such price fluctuations, especially for the timeframes less than one hour (H1).

Try to show the problem area in more detail, please. The attached picture is not very clear.

 
Renat:
The tester starts testing at 100 bars from the beginning to avoid potential errors when working with an empty history. Therefore, the test does not start at 00:00.

Most likely, it will not be empty and you can check it, and the tester should work accordingly (if it is empty, it steps back 100 bars, if it is not, it starts from 00:00).
OK. Suppose from the 100th bar from the beginning - timeframe 5 minutes, 100 bars = 500 minutes / 60 ~ 8.3 hours. And the test starts after 2 hours. It doesn't add up either.
Where does the last hour go
? The tester finishes at 23:00 at the limit before the current date. Why not at 23:59:59 ? The last hour is not superfluous at all )))

Renat:

Try to show the problem area in more detail, please. The attached picture is not very clear.

Expert Advisor has OnTick() handler of the following form:

void OnTick()
{
double ask = SymbolInfoDouble("EURCHF", SYMBOL_ASK);
double bid = SymbolInfoDouble("EURCHF", SYMBOL_BID);

Print("ask: ", ask, " bid: ", bid);
}

During a test run on the EURUSD chart we get different quotes from a test run on the EURCHF chart (for example),
Although the range and timeframe of the test, as well as the symbol for which we receive quotes SymbolInfoDouble("EURCHF",...) don't change.
Thus, it turns out that the results of a test varies depending on the symbol selected. They should be the same.



 
Renat:
The result depends entirely on the examiner himself and his sensitivity to price flow. The results are most accurate when tested on a tickwise basis.
It seems to me that this phenomenon is possible if in calculations the expert uses the current bar that is forming.
 

crOss, the thing is that if you choose different tools when testing the multicurrency, the results may be different if you use OnTick. Especially when you run the Expert Advisor on open prices.

The bar of one instrument always opens before the bar of another instrument, because the ticks always follow the same flow. And a tick for one instrument will always be earlier than a tick for another instrument, even if they are marked with the same time. By the way, a bar can open not exactly at 0 seconds, but in some seconds or even minutes.

That is, if you test on the opening of a bar of one instrument, the corresponding bar of another instrument has not yet opened. And if you test the opening of a bar of another symbol, the corresponding bar of the first instrument has already opened.

That's why there can be a difference

Документация по MQL5: Стандартные константы, перечисления и структуры / Константы индикаторов / Ценовые константы
Документация по MQL5: Стандартные константы, перечисления и структуры / Константы индикаторов / Ценовые константы
  • www.mql5.com
Стандартные константы, перечисления и структуры / Константы индикаторов / Ценовые константы - Документация по MQL5