Strategy Tester is nearly unusable...

To add comments, please log in or register
Luciano Fagundes
25
Luciano Fagundes  

Hi Folks


I have found many issues with Strategy Tester over the last couple days. Could you please tell me if there is anything I can do to fix any of them? Here they are : 


1) Start Time : I can set a start and end date for the testing, however, the process starts at 00:00 of the day I selected. This is terrible. Consider that the business day in Brazil starts at 09:00 AM and the system process "nothing" from 00:00 to 09:00. It means that it takes almost 10 minutes doing "nothing" until some useful data start showing. Can I tell the tester to process only business hours?

2) There are assets that most of the deals take place in a single month. For instance currency exchange between US$ and BRL. Those contracts show up in a continuous series and on month-based series. When I run the tester with a continuous series, it blocks me from generating deals. I get an error messages saying that volume is invalid. I'd guess that it does not allow me to buy assets from the continuous series. However, when I use the real asset symbol, some times the tester work and some times it does not, see the next item...

3) Tester stops right before the day will begin. As I said, it starts processing at 00:00. Several times, when it arrives as 08:59, the processing finishes. There are no error or anything like that. It just seems to finish as it it had completed the testing. This usually happens when I set a custom period and execution as "1 minute OHLC". After I reboot the computer, some times, the system process up to the end of the first or second day. But it has never passed the second day ( I'd like to run it for a full week... ).

4) It slows down exponentially. When the tester is doing "nothing" from 00:00 to 09:00, it goes by really fast. Then, at 09:00, it starts processing in a nice rate, I'd guess about a couple minutes per second. However, as the processing goes by, it slows down a lot. To the point that I end up stopping the testing because it gets slower than real time ( more than a minute for each minute ). And, there is not good reason for that because I see on my computer that it is taking about 300MB of RAM and using only about 4% of the CPU power. It must be something related to the tick generator that slows down for some reason.


So, is this familiar to any of you ? Any suggestions on how to improve the strategy tester performance ? I'm afraid that if this does not work I will probably need to migrate to a different platform...


Thanks


-Luciano  

Stuart Browne
7440
Stuart Browne  

As per your thread title, YES!

Personally, I use strategy tester only for logic testing an EA. Profit testing is only useful in live/demo forward testing. Besides the 4 points you listed above, there are quite a few more "features" that make strat test in MT useless as a profit predictor

Stanislav Korotky
24522
Stanislav Korotky  

1) Use input parameters in your EA to setup hourly session frames for trading.

2) and 3) are unclear for me.

4) Looks like an issue in your code.

Luciano Fagundes
25
Luciano Fagundes  
Stanislav Korotky:

1) Use input parameters in your EA to setup hourly session frames for trading.

2) and 3) are unclear for me.

4) Looks like an issue in your code.


for 1) I do that. My EA does nothing after hours, however, the tester goes all the way from 00:00 up to 09:00, this alone takes about 10 mins before the test actually begin...

2) is related to special assets...

3) This is one of the oddest ones. The tester goes from 00:00 up to 08:59 and when the day is about to start, the tester finishes. No errors, no nothing, just says that the testing is completed.

4) Not sure if this is an issue in my code, it runs well, no CPU is used, Memory is not leaking, I'm trying to figure that out but the only thing I see is that the ticks are coming really slow....


Thanks 

Luciano Fagundes
25
Luciano Fagundes  
Stuart Browne:

As per your thread title, YES!

Personally, I use strategy tester only for logic testing an EA. Profit testing is only useful in live/demo forward testing. Besides the 4 points you listed above, there are quite a few more "features" that make strat test in MT useless as a profit predictor


I'm almost considering that too.. I'd like to run the EA for a few days and see how it goes. Do you know of any other tools that would allow me to do that ? Many people here like chart, is it any good? I see they have a completely different programming language, I'd need to rewrite my entire code for that so I'm avoiding the idea for now... 

Enrique Dangeroux
408
Enrique Dangeroux  
Luciano Fagundes:

I'm almost considering that too.. I'd like to run the EA for a few days and see how it goes. Do you know of any other tools that would allow me to do that ? Many people here like chart, is it any good? I see they have a completely different programming language, I'd need to rewrite my entire code for that so I'm avoiding the idea for now... 


I am sure it has nothing to do with tester, but with your code. Try it on a demo account for a day, it will probably behave the same.

On a side note. Testing 1 week will yield no decision making data. Testing with MT4 ditto.

Alain Verleyen
36033
Alain Verleyen  
Luciano Fagundes:

Hi Folks


I have found many issues with Strategy Tester over the last couple days. Could you please tell me if there is anything I can do to fix any of them? Here they are : 


1) Start Time : I can set a start and end date for the testing, however, the process starts at 00:00 of the day I selected. This is terrible. Consider that the business day in Brazil starts at 09:00 AM and the system process "nothing" from 00:00 to 09:00. It means that it takes almost 10 minutes doing "nothing" until some useful data start showing. Can I tell the tester to process only business hours?

Right you can't set a starting hour. Though you can easily manage that in your EA (see Stanislav answer). If it takes 10 minutes, then you have a bug in your code or a bad logic.

2) There are assets that most of the deals take place in a single month. For instance currency exchange between US$ and BRL. Those contracts show up in a continuous series and on month-based series. When I run the tester with a continuous series, it blocks me from generating deals. I get an error messages saying that volume is invalid. I'd guess that it does not allow me to buy assets from the continuous series. However, when I use the real asset symbol, some times the tester work and some times it does not, see the next item...

Bug in your code.

3) Tester stops right before the day will begin. As I said, it starts processing at 00:00. Several times, when it arrives as 08:59, the processing finishes. There are no error or anything like that. It just seems to finish as it it had completed the testing. This usually happens when I set a custom period and execution as "1 minute OHLC". After I reboot the computer, some times, the system process up to the end of the first or second day. But it has never passed the second day ( I'd like to run it for a full week... ).

Bug in your code.

4) It slows down exponentially. When the tester is doing "nothing" from 00:00 to 09:00, it goes by really fast. Then, at 09:00, it starts processing in a nice rate, I'd guess about a couple minutes per second. However, as the processing goes by, it slows down a lot. To the point that I end up stopping the testing because it gets slower than real time ( more than a minute for each minute ). And, there is not good reason for that because I see on my computer that it is taking about 300MB of RAM and using only about 4% of the CPU power. It must be something related to the tick generator that slows down for some reason.

Bug in your code.

So, is this familiar to any of you ? Any suggestions on how to improve the strategy tester performance ? I'm afraid that if this does not work I will probably need to migrate to a different platform... 

Fix the bugs.

Alain Verleyen
36033
Alain Verleyen  
Stuart Browne:

As per your thread title, YES!

Personally, I use strategy tester only for logic testing an EA. Profit testing is only useful in live/demo forward testing. Besides the 4 points you listed above, there are quite a few more "features" that make strat test in MT useless as a profit predictor

You are mixing completely different things Stuart.

Globally MT5 tester is working well, as expected and documented. They are some bugs in specific situations, these bugs if important are fixed relatively fast. Of course it could always be improved, but that's an other point too.

If the Strategy Tester is usable to provide reliable trading results, is an other matter (already discussed numerous times on this forum).

Stanislav Korotky
24522
Stanislav Korotky  
Luciano Fagundes:

for 1) I do that. My EA does nothing after hours, however, the tester goes all the way from 00:00 up to 09:00, this alone takes about 10 mins before the test actually begin...

2) is related to special assets...

3) This is one of the oddest ones. The tester goes from 00:00 up to 08:59 and when the day is about to start, the tester finishes. No errors, no nothing, just says that the testing is completed.

4) Not sure if this is an issue in my code, it runs well, no CPU is used, Memory is not leaking, I'm trying to figure that out but the only thing I see is that the ticks are coming really slow....


I think nobody can help you until you provide more details about the tester settings, work instruments/timeframes, and some info about EA. Feel free to share some screenshots, logs, etc.

The tester is indeed not so convenient as it could be, but the problems you described seems having their origin in some other things.

Marco vd Heijden
Moderator
8151
Marco vd Heijden  

I just stopped using it, which solved the problem.

Stuart Browne
7440
Stuart Browne  
Alain Verleyen:

You are mixing completely different things Stuart.



Sorry it confused you Alain. Sounds like a few others, including the OP, know exactly what I'm talking about

12
To add comments, please log in or register