problem - different results with optimization with remote agents (mql5 cloud) VS local agents/tester
it was issue with local agents that disapeared on intself... computer restart did not help. Changing the test period to longer (2012.01 instead 2013.01) then back to the same and changing testing mode (1H opening prices then back to 1m opening prices again) solved the issue.
Maybe that GBP/NZD pair is cursed as I encounter more and more problems while trying to optimize settings for it. ;)
Did anyone encounter problems like that?
PS. the problem was that i was using SMA 440 and started to test at 2013.01. I dont know why - but the remote agents produced different results than the local agents. When i extended the back-test to 2012.06 the problem was solved and both local agents and remote agents acted the same. The strategy has build-in mechanism for fixed startup period (here 2013.06). The tester does not provide enough history for large-number MAs. Watch out for that.
PS2. There is still little difference in results (i have to increase money management balance by about 10% not to get equity wipe-out in comparison to MQL5 cloud results but its nothing I cant handle)
the problem is still there. I managed to recreate the problem, then fix it, then recreate it again, then again fix it. All on the attached AVI clip.
It should not work like this!
the problem is still there. I managed to recreate the problem, then fix it, then recreate it again, then again fix it. All on the attached AVI clip.
It should not work like this!
i can just do it with this certain settings with my ea. I don't know why it does occur though and if i can replicate it in any other situation.
Funny thing is: i run test - its ok and profitable. Running same exact settings and everything again (just pushing start button again) and equity wipeout. I change date again by 1 day and start - it is ok, then i press start again - equity wipeout. :o The problem always occurs when i test same settings the second time.
I checked on other copy of MT5 and its fine. What file i should delete/replace to fix the issue?
I already reinstalled the terminal and replaced Config/agents.dat and the issue still occurs.
I have the same problem.
The thing is that float variables are just approximations of real numbers, and that happen for one obvious reason: real numbers are infinite and a float variable has only 32 bits to store it. Because of the difference in floating-point hardware available on different architectures the result may be slite different.
Those differences are really (really) small, but If you start to add up this small errors over and over, at some point it will make a difference.
I don't have a solution for you, but I hope this helps you to at least understand why it happens.
Indeed if the resolution between them is different it can hit certain values on one where it can miss on the other it's more of a quality kind of thing like megapixels on an image? . sure we would like to see it making no difference in result but the only thing that will give you real results is running it live by throwing some money at it, and if i look at your profile you seem to have plenty :)
Also you can try if the simulation speed has got anything to do with it, if you run it too fast it can't keep up with calculating and start to generate errors and bad sectors in the data so i think its a good idea to run it slower and see if it still happens.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Like in the topic. Sometimes (like now) i have issues redgarding different results in paid agents optimization and results from my local agents.
How is that possible? Its only eating my credits and producing useless results.
Attaching screens of remote paid agent optimizations (mql5 cloud) and local agents optimization with the same settings.
The first screenshot is MQL5 cloud optimization.
Second is the same optimization but on local agents.
PS. I noticed some problems with it in the past but never that obvious.