On-line trading advisor. Exchange of views - page 12

 
OnGoing:

More. You write on the website that "working in the tester almost brought you to a dead end". However, this is the reason why there are great doubts about the sustainability of your strategy. Since some mythical "internal self-testing" may be nothing more than self-deception.

I don't trust the tester either. However, if an expert does not stand up to at least a rough test in a tester, I personally would not rely on the test data on a forward, even over a year. For they can be very static, i.e. depend on the coincidence of external factors. In the tester, they are cut off, and all that is left is the backbone of quotes and nothing else. Yes, there can be gaps and inconsistencies in the history, but nevertheless with some analysis and diligence they can be easily found, and the results obtained due to them will be excluded from the general picture.

All the more so, a grid based on purely mathematical models simply has to undergo testing. Otherwise, it is possible to remain in illusions about its effectiveness for years. And at the next unaccounted "force majeure", again and again "stalemate", waiting for months to the next similar case. Quite an irrational testing model you have chosen, that's my opinion.

Yes, there is actually a lot of time spent testing new ideas and changes. The very backbone of the idea was originally tested in the strategy tester. It is just not clear what exactly needs to be tested: the parameters are all calculated in automatic mode, and I do not enter anything manually, so I can check it in the form of a changing parameter.

I tried to run it in the strategy tester. The result gives out:

2011.09.13 15:41:13 TestGenerator: unmatched data error (high value 1.4321 at 2009.12.31 18:43 and price 1.4321 mismatched).

In short, not a single transaction.

Also, there are messages:

2011.09.13 15:44:44 2009.01.02 06:01 Lakshmi_4_294 EURUSD,M1: stack overflow
2011.09.13 15:44:44 2009.01.02 06:01 Lakshmi_4_294 EURUSD,M1: 1002

Apparently, it is trying to download historical data and stops at one of the missing bars.

 
forexnew:

Yes, it actually takes a lot of time to test new ideas and changes. The backbone of the idea itself was originally tested in the strategy tester. It's just not clear what exactly needs to be tested: the parameters are all calculated in automatic mode, and I don't manually enter anything to be able to test as a changing parameter.

I tried to run it in the strategy tester. The result gives out:

2011.09.13 15:41:13 TestGenerator: unmatched data error (high value 1.4321 at 2009.12.31 18:43 and price 1.4321 mismatched).

In short, not a single transaction.

Also, there are messages:

2011.09.13 15:44:44 2009.01.02 06:01 Lakshmi_4_294 EURUSD,M1: stack overflow
2011.09.13 15:44:44 2009.01.02 06:01 Lakshmi_4_294 EURUSD,M1: 1002

Apparently, it is trying to download historical data and stops at one of the missing bars.

Probably the terminal settings just did not set a higher value for the permissible number of bars on the history.
 
OnGoing:
Probably, the terminal settings just did not specify a larger value of the permissible number of bars on the history.

Yes, it is:

Max history bars 10000000

Max bars in window 10000

 
It is also possible that the switching between timeframes is glitchy. The Expert Advisor switches between timeframes while downloading the historical data. And the strategy tester is set to one timeframe. This is only a guess. The Expert Advisor stopped being tested at the beginning of version 4 and now it is version 4.294.
 
I can look it up, who knows, maybe two heads are better than three in this case)
 

The first thing to do is to check that all values in all actions are normalised in both data and conditions. This will remove a good half of the errors, if not more...

 
forexnew:

Yes, it actually takes a lot of time to test new ideas and changes. The backbone of the idea itself was originally tested in the strategy tester. It's just not clear what exactly needs to be tested: the parameters are all calculated in automatic mode, and I don't manually enter anything to be able to test as a changing parameter.

I tried to run it in the strategy tester. The result gives out:

2011.09.13 15:41:13 TestGenerator: unmatched data error (high value 1.4321 at 2009.12.31 18:43 and price 1.4321 mismatched).

In short, not a single transaction.

Also, there are messages:

2011.09.13 15:44:44 2009.01.02 06:01 Lakshmi_4_294 EURUSD,M1: stack overflow
2011.09.13 15:44:44 2009.01.02 06:01 Lakshmi_4_294 EURUSD,M1: 1002

Apparently, it is trying to download historical data and stops at one of the missing bars.



I don't think so - rather, the program is trying to put some information (reference) in the same place very many times. Stack overflow happens either by abusing recursive subroutine calls or by using some data types incorrectly.
 
tara:

I don't think so - rather, the program is trying to put some information (reference) in the same place very many times. Stack overflow occurs either by abusing recursive subroutine calls or by using some data types incorrectly.
Digging through more than 4 thousand code lines will be troublesome. You should probably take old versions of the program and see when this error appeared.
 
There will be an on-line webinar today at 6pm in Ukraine (19pm in Moscow). If anyone has any questions, you are welcome to attend. I feel like a schoolboy at an exam :))
 
forexnew:
There will be an on-line webinar today at 6pm in Ukraine (19pm in Moscow). If anyone has any questions, you are welcome to attend. I feel like a schoolboy at an exam :))

Share the link...