Why the developers failed to make a proper tester - page 3
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
It has nothing to do with normalisation... My price is a number prescribed in externa's variable types and is clearly in this case 1.9850, without a fifth sign!!!!
Then, it must be due to unmasking errors.
It's true, there are pitfalls, but you can do it! But the speed of testing and optimization is dozens of times slower!
Already found it, MT4, MT5, java, C! I'm shocked... and the options for pitching the story! Tics, candles, splines!!! I'm almost in love! I do not want to say bad things about MT, very decent and interesting platform, but it lacks a lot, and here I have everything at hand, although these are first impressions, just a few hours of digging...
What about the pitfalls, can you tell me more about it?
So write on mt4 for jforex ! True, there are pitfalls, but it can be done! But the speed of testing and optimization is dozens of times slower
Then it must have been a deflagration error.
Right, that must be it...
I came to the conclusion a long time ago that the tester cannot be trusted at all! I'm not even talking about testing for all ticks and optimisation, but even opening prices in different test modes produce different results. Here is an example, two tests, the only difference is that "visualization" is enabled in one of them and disabled in the other:
The results have nothing in common.
And in MT4 there are four components that you can really use only the terminal and MQL, the tester (+optimizer) and the eternally falling down editor for x64 should not be touched.
Had to use this editor briefly today, cussed it out, crashes on copy-paste every once in a while, Win7 x64. The problem with x64 is well known.
I've been writing to developers for a long time, they don't give a damn.
Man, on the English-speaking part of the forum they were just harshly berating them like "Russian Ivans still don't know that there are 64-bit OSes", I couldn't stand it, corrected the bug. Whatever... DC pays the money,
and the DC doesn't need an editor.
Damn it again, Microsoft specially wrote manuals on how to adapt programs for Vista, Vyn 7 and 64-bit systems, they even have them in Russian, I think! It's not so hard, just read it and do it!
Why don't my programs crash on x64? After all, you don't have to try hard at all, just read the instructions and make corrections!
And you are scolding the tester... Anything that is not needed by DC is second-rate by definition.
ZS: But the tester really should be written for itself, and the tick history should be written for itself, what I've been doing for a long time for three DCs
I came to the conclusion a long time ago that the tester cannot be trusted at all! I'm not even talking about testing for all ticks and optimisation, but even opening prices in different test modes produce different results. Here's an example, two tests, the only difference is that in one the "visualisation" checkbox is on and in the other it is off:
Angela, check again. The tester takes information about the spread directly from the current market, so it happens that when the spread is temporarily widened, the tester starts to underestimate the values inadequately. Your case may well be due to this. Although I'm not excusing MQ, this situation could have been foreseen (at least the possibility to manually set spreads).
I don't quite understand you, I work in autonomy, I do two consecutive runs, only switching the visualization checkbox, the story in the tester is the same, including information about spreads.
The only conclusion from this, which I raised in other threads before, is that tester's algorithms work differently in different modes, and some situations are handled differently. I can assume in this particular case, because I use global variables, processing of global variables in visualization mode and without it is different, I can't even imagine anything else. As for checking again, I've raced many times in one and the other mode, the results are the same.
I don't quite understand you, I work in autonomous mode, I do two consecutive runs, only switching the visualization checkbox, the story in the tester is the same, including information about spreads.
The only conclusion from this, which I raised in other threads before, is that tester's algorithms work differently in different modes, and some situations are handled differently. I can assume in this particular case, because I use global variables, processing of global variables in visualization mode and without it is different, I can't even imagine anything else. As for checking again, I've raced many times in one and the other mode, the results are the same.