Tester in new build 604 doesn't work correct - page 8
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
What size is your .fxt file ? if it's bigger than 4GB then that is the cause.
We cannot get our complex, live, real-money EAs back-testing properly (i.e. achieve the same test results with 60x builds that we were getting with 509 and prior). They appear to trade properly (though it's not so easy to tell the detail), but back-testing shows very different results. There are so many different potential problems causing this, it's been impossible so far to work out which ones are the culprits and very difficult to fully eliminate anything from the list.
My post doesn't add any technical value I'm afraid but it does demonstrate that there are still capable people/teams out there struggling away with the new reality.
It shouldnt be that big, my 4yr backtests generated fxt files of about 2.75 GB
Hi SDC,
Given the changes in respect to historical data, that is one of the areas we've been looking at. We've been trying to eliminate it from the list of potential problems.
I had the following (edited) interaction with one of the MQ guys yesterday (bearing in mind we sourced/paid for all our M1 data directly from the broker and analysed it thoroughly for completeness before use and have not altered it since):
Trev: We have always used the strategy tester in a live, broker-connected terminal where the M1 .hst files are being continually added to with new M1 broker data. I guess my question is, was anything done to my existing M1 .hst file as part of the terminal upgrade, and what is happening to that file as the new data comes in (i.e. is it being updated in the correct format)?
MQ: In the new MT4 600 for testing experts do not need do anything.
The answer isn't really what I was looking for (though it may be just semantics). I want to know the details around what has been done to my .hst's (if anything), as part of the upgrade. Has the whole thing been converted to the new format such that the older data is already usable with the new terminal's tester, and is the new data being fed in going in in the right way? If the answer to those question is yes (and assuming it's been done accurately), then I can answer your question SDC with 100% affirmation and remove it from my issues list.
Hi SDC,
Given the changes in respect to historical data, that is one of the areas we've been looking at. We've been trying to eliminate it from the list of potential problems.
I had the following (edited) interaction with one of the MQ guys yesterday (bearing in mind we sourced/paid for all our M1 data directly from the broker and analysed it thoroughly for completeness before use and have not altered it since):
Trev: We have always used the strategy tester in a live, broker-connected terminal where the M1 .hst files are being continually added to with new M1 broker data. I guess my question is, was anything done to my existing M1 .hst file as part of the terminal upgrade, and what is happening to that file as the new data comes in (i.e. is it being updated in the correct format)?
MQ: In the new MT4 600 for testing experts do not need do anything.
The answer isn't really what I was looking for (though it may be just semantics). I want to know the details around what has been done to my .hst's (if anything), as part of the upgrade. Has the whole thing been converted to the new format such that the older data is already usable with the new terminal's tester, and is the new data being fed in going in in the right way? If the answer to those question is yes (and assuming it's been done accurately), then I can answer your question SDC with 100% affirmation and remove it from my issues list.
The data are the same, only the .hst format was changed. As SDC's test shows, backtesting results are the same with both build 509 and 60x. But it also shows that SDC first attempt was corrupted because he doesn't use the same data.
The broker can change your data though.
I can see you are using a new "events", try use old mt4 syntax. for 60* versions.
thanks,regards
I did that, it made no difference to the result whether I used start() or OnTick() in the new builds. I decided to use the new syntax and #property strict in the new compiler to maximize the potential for difference.
While I still have both builds with identical chart data, anyone who beleives they have identifiied a problem that changes the way their code runs in the new builds is welcome to modify the test code to include an example of that issue, I will then run the tests again so we can see what effect it has on backtesting.
While doing these tests I did notice a couple of minor bugs.
1) In the new terminal when logging in and out of different accounts, the little man in the navigator stays green on all of them, making it difficult to tell which account is logged in.
2) The spread box in strategy tester. When I enter the spread as 3, The value in that box changes to 30 by itself intermittently. This may have something to do with 5 digit account .... is it trying to tell me I should have entered 30 because it is 5 digits ?
OK I added an init() to deinit() elapsed time calculation.
1yr back test, EUR/USD, 1 minute charts
Build 509
Build 606
Build 606 strategy tester was on average 11.8% faster then that of build 509
Was this visual or non visual mode ?
non visual.
ok at 32x for the whole test ? I am running these tests on a ten year old AMD 2.4 Ghz duel core laptop so dont expect to be amazed by speed tests lol