Errors, bugs, questions - page 488

[Ilya Belyaev]

Bild 489. 32x. Noticed a bad feature of the tester here... the last hour drops out of the test range,
if the range is closed with the current date. For systems that work on small timeframes this hour may be important.
It would be nice if we could set the range boundary at the current time of the test start.
This would increase the relevance of parameters when optimising trading systems and consequently improve the performance.

[Lyuk]

It seems to be a bug when running the tester on a minute timeframe and a large number of trades (almost every bar), actually it should be a straight line diagonally without sharp breaks at the end.

I tried to ask the Expert Advisor why it abruptly loses at the end of any historical data and found out that it loses evenly :), it is just that this chart is compressed at the end along the date axis, i.e. in the last centimeter of the chart there are more than 10 days, while the rest part has only 3 days.

I'm not sure exactly (maybe I messed up something in the EA), please check it.

[Igor Volodin]

Lyuk:
Kind of a bug when running the tester on a minute timeframe and a large number of trades

It's definitely a bug. I've also left a request in the SD

[Vladimir Gomonov]

Lyuk:

1) ....... It is just that this graph is compressed on the date axis at the end, i.e. in the last centimetre of the graph - more than 10 days, whereas the rest of the graph is only 3 days.

2) I'm not sure exactly (maybe I messed up something in my Expert Advisor), please check it.

1. The end of the chart is compressed with a large number of trades.

2. I have not, it is exactly the way it is. This bug has been circulating from build to build for a long time now. I hope it will be fixed someday.

[Igor Volodin]

crOss:

Bild 489. 32x. Noticed a bad feature of the tester here... The last hour is dropping out of the test range

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

[Olegs Kucerenko]

Hello, I have one expert working better with ticks in my tester than with OHLC. The other one shows miracles even though I have short trades but average ones is more than 30 pips. During testing the number of deals is the same, graphic layout is the same. The calculations may be different. Results are surely diametrically opposed :-) It's interesting that the initial period is almost coincident graphically... Some trades are better, some are worse... But further ones are absolutely different... Besides sometimes I fail mightily... Displaying a large number of trades is cool, but I choose duration of deals so this bug will not reveal itself.

[Aleksey Lebedev]

The "Experts" log skips every 10th print, i.e. the script above does not print numbers ending in nine. All is normal in the logs.

//+------------------------------------------------------------------+
//|                                                         Test.mq5 |
//+------------------------------------------------------------------+
//| Script program start function                                    |
//+------------------------------------------------------------------+
void OnStart()
  {
   for(int i=0;i<100;i++) Print("i=",i);
  }
//+------------------------------------------------------------------+

[Renat Fatkhullin]

Karlson:

Hello, I have one expert working better with ticks in my tester than with OHLC. The other one shows miracles even though I have short trades but average ones is more than 30 pips. During testing the number of deals is the same, graphic layout is the same. The calculations may be different. Results are surely diametrically opposed :-) It's interesting that the initial period is almost the same in real time: some trades are better, others worse... But further ones are absolutely different... Sometimes I fail too much. I know how many trades are displayed, but I choose duration of trades so this bug will not reveal itself.

The result is entirely dependent on the Expert Advisor and its sensitivity to price flow. The results are most accurate when tested on a tick-test.

[Renat Fatkhullin]

Swan:

The "Experts" log skips every 10th print, i.e. the script above does not print numbers ending in nine. All is normal in the logs.

Output of logs to screen works with output frequency cutoff, so as not to slow down on massive printers. Everything is written to disk without change.

[Aleksey Lebedev]

Renat:
Log output to screen works with output frequency cut-off so as not to slow down on bulk printers. Everything is written to disk unchanged.

The achievable effect of the cutoff is questionable, the number of prints decreases by 10%, and for full information it is necessary to get into the log...

and it's not a tester, so there's no hurry.)

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