Tester giving identical results even with a factor of 10 difference in spread Why? - page 2

 
Lorentzos Roussos #:

Both tests use the spread specified in the rates yeah

I would suggest the op did a test with new symbol names as he may be stepping on another symbol possibly (i.e. new bars did not import or smth)

The posted log show a variable spread between 3 and 7 points (quick checking) and the same on both tests.

Ok. Understood. The EA is not trading the custom symbols but the broker symbols.

CS    0    18:55:57.044    Multi Symbol EA Master (AUDCHF.custom.spread=200,M15)    2020.06.18 02:45:05   1, AUDCHF, 0.100000, 0, 0.652230, 0.000000, 0.000000
...

Trade    2020.06.18 02:45:05   market buy 0.1 AUDCHF (0.65210 / 0.65223 / 0.65210)
CS    0    18:55:57.152    Trades    2020.06.18 02:45:05   deal #2 buy 0.1 AUDCHF at 0.65223 done (based on order #2)
CS    0    18:55:57.152    Trade    2020.06.18 02:45:05   deal performed [#2 buy 0.1 AUDCHF at 0.65223]
CS    0    18:55:57.152    Trade    2020.06.18 02:45:05   order performed buy 0.1 at 0.65223 [#2 buy 0.1 AUDCHF at 0.65223]

 
EdFuk #:
Thanks for looking into it. It’s very strange. I’m planning to spend a bit of time on it tomorrow. 

Before going down the custom symbol route I tried to change the spread field in the symbol to an integer. That didn’t seem to work but should it?

from the tickstory data ? maybe

May i ask what the import process you used is ? 

My terminal does not let me add bars by hand to custom symbols 

 
Alain Verleyen #:

The posted log show a variable spread between 3 and 7 points (quick checking) and the same on both tests.

Ok. Understood. The EA is not trading the custom symbols but the broker symbols.

CS    0    18:55:57.044    Multi Symbol EA Master (AUDCHF.custom.spread=200,M15)    2020.06.18 02:45:05   1, AUDCHF, 0.100000, 0, 0.652230, 0.000000, 0.000000
...

Trade    2020.06.18 02:45:05   market buy 0.1 AUDCHF (0.65210 / 0.65223 / 0.65210)
CS    0    18:55:57.152    Trades    2020.06.18 02:45:05   deal #2 buy 0.1 AUDCHF at 0.65223 done (based on order #2)
CS    0    18:55:57.152    Trade    2020.06.18 02:45:05   deal performed [#2 buy 0.1 AUDCHF at 0.65223]
CS    0    18:55:57.152    Trade    2020.06.18 02:45:05   order performed buy 0.1 at 0.65223 [#2 buy 0.1 AUDCHF at 0.65223]

Doh!!


Thanks Alain. Yes you got it. I forgot to change the traded symbol in the EA. FFS!


Thanks for your help.

 
Lorentzos Roussos #:

from the tickstory data ? maybe

May i ask what the import process you used is ? 

My terminal does not let me add bars by hand to custom symbols 

Hi Lorentzos. Go to the symbol manager and set the custom symbol specification. I just based mine on the broker AUDCHF. Then go to the bars tab. Make sure the custom symbol is selected in the combo box then click import. Another dialog appears. Browse for the tickstory file you have downloaded. Select it and the data should appear in the table. Then click OK and it will start to import the data


 

@Alain Verleyen

I did some tests by modifying the spread integer in the custom symbol as shown. Doesn't appear to be working. Should it be? Would be much easier than having to modify the large data files.



 
EdFuk #:

Hi Lorentzos. Go to the symbol manager and set the custom symbol specification. I just based mine on the broker AUDCHF. Then go to the bars tab. Make sure the custom symbol is selected in the combo box then click import. Another dialog appears. Browse for the tickstory file you have downloaded. Select it and the data should appear in the table. Then click OK and it will start to import the data


Thank you that's how i do it but with my csv , perhaps i'm doing something wrong on the export .(of the csv)

EdFuk #:

@Alain Verleyen

I did some tests by modifying the spread integer in the custom symbol as shown. Doesn't appear to be working. Should it be? Would be much easier than having to modify the large data files.



How did you adjust the tickstory files , excel ? If the structure is known maybe you could fashion a tool to quickly do that for you (although excel would not be slow either) 

 
Lorentzos Roussos #:

Thank you that's how i do it but with my csv , perhaps i'm doing something wrong on the export .(of the csv)

How did you adjust the tickstory files , excel ? If the structure is known maybe you could fashion a tool to quickly do that for you (although excel would not be slow either) 

I wrote a script in Python to do it. I’m currently doing a sensitivity study on the effect of spread on my test results. The only way of getting accurate results is to use real ticks however if possible I’d like to use 1min OHLC for a preliminary run. At the moment I am seeing big differences and trying to drill down to exactly why. I will post my findings. 
 
EdFuk #:
I wrote a script in Python to do it. I’m currently doing a sensitivity study on the effect of spread on my test results. The only way of getting accurate results is to use real ticks however if possible I’d like to use 1min OHLC for a preliminary run. At the moment I am seeing big differences and trying to drill down to exactly why. I will post my findings. 

I see , so you will do a "pessimistic" run with a wider than real spread and that solution then you'll let it optimize on real ticks 

 

Well I've just been doing some testing and experienced some very strange behaviour.

I was expecting 1 min OHLC and Open Prices to give very similar results as I had created two custom symbols, both with a spread of 1 point. What I got was two curves that diverged from each other (The red line and green line in the figure below). I spent ages trying to figure out why and debugged through the code, checked the input data. The input data was fine and when I debugged I found that the spread when running the 1min OHLC case was 24....curious!! as I have another custom symbol that I forced the spread to be 24. I deleted the cache but that didn't seem to make a difference. Then I deleted the symbol and recreated it with a different name and ran it and got the blue line below which is very similar to the red and what I was expecting.


Can anyone explain what was happening? I checked the symbol and it displayed the correct data in the viewer but was obviously picking up data from elsewhere or previous? I did create a custom symbol based on another one but I thought that just pulls across the specification and not a link to the data etc?

I will try and replicate it.

It's making me very nervous that it's ease to do analysis with incorrect data.


**further update. I deleted the symbol named "AUDCHF.custom.spread=1" from the symbol list, shut down MT5 then reopened it, created a new symbol with the same name as before and it gave the same (incorrect) results. Still picking up the file with a spread of 24 rather than the spread=1.

If you create a symbol with a completely unique name that hasn't been used before it works correctly. Seems like there's a pointer to an old file somewhere. Any ideas?




 
EdFuk #:

Well I've just been doing some testing and experienced some very strange behaviour.

I was expecting 1 min OHLC and Open Prices to give very similar results as I had created two custom symbols, both with a spread of 1 point. What I got was two curves that diverged from each other (The red line and green line in the figure below). I spent ages trying to figure out why and debugged through the code, checked the input data. The input data was fine and when I debugged I found that the spread when running the 1min OHLC case was 24....curious!! as I have another custom symbol that I forced the spread to be 24. I deleted the cache but that didn't seem to make a difference. Then I deleted the symbol and recreated it with a different name and ran it and got the blue line below which is very similar to the red and what I was expecting.


Can anyone explain what was happening? I checked the symbol and it displayed the correct data in the viewer but was obviously picking up data from elsewhere or previous? I did create a custom symbol based on another one but I thought that just pulls across the specification and not a link to the data etc?

I will try and replicate it.

It's making me very nervous that it's ease to do analysis with incorrect data.


**further update. I deleted the symbol named "AUDCHF.custom.spread=1" from the symbol list, shut down MT5 then reopened it, created a new symbol with the same name as before and it gave the same (incorrect) results. Still picking up the file with a spread of 24 rather than the spread=1.

If you create a symbol with a completely unique name that hasn't been used before it works correctly. Seems like there's a pointer to an old file somewhere. Any ideas?




are you using the tester_file property(if i recall its name properly) maybe ? 

the Tester file when you change it , the file name or the file contents , you must do a compilation + close ide + mt5 + restart them and then test

i recall losing (wasting) ages too on that so it sounds familiar