Open project - tester-optimiser in-house - page 5

 
Every vegetable has its due date. As soon as I'm ripe, I'll post it. There's a lot of people walking around, maybe someone will ripen sooner...
 
The tester should not be written in McLean, but in some high-level language. I'm not sure that something will change dramatically in MT3, and I'm not sure that in MT4. I've just written a preparation for one strategy in Delphi, like i.e. optimal parameters selection by gradient descent method. Although there was only a prototype of the tester (working with one order only, no pending orders), it was succeeded 1000 times in less than 2 seconds for 11000 bars, besides the strategy was not bad - a trend on three TFs was taken into account. And the built-in scripts allow writing a tester in MT only out of complete hopelessness. By the way, I tried to implement all described above in MT3, but there is such a thing as LupDetect and I finally gave up using MT. The more so, since any high-level language gives me more possibilities (if not two orders of magnitude).
By the way, a question - does McLe4 support object-oriented programming?
 
First. OOP is not supported in mql-4, neither inheritance nor methods or properties, in general there are no objects or classes.
The second one. In MT4 we can easily receive values for different TFs, even three, even all 9 from a minute to a monza.
Third. МТ4 will not be able to detect. It was removed.
Fourth. The productivity was increased by 40 times. For more details see here - "MQL4, MQL2, EasyLanguage, Wealth-Lab 3.0 and VC++: Speed Comparison".
 
Will MT4 support non-standard TFs? For example, M90 or M45 or D2?
Tests show that standard TFs are not always the most effective.
 
No, it won't. This point from the developers is ironclad. Many things have been penetrated by beta-testers, but here there is an ironclad level of resistance. I think in MT5 there will be not only TFs of any kind, but also RadarScreen and many other things...
 
URAN wrote
I don't understand how to use it, post an example with some experts, if it's not difficult... <br / translate="no">



Here I made an example. It will not work. Source here - http://forexsystems.ru/phpBB/viewtopic.php?p=4636#4636
 
Judging by how much MT4 has been made, it is premature to talk about MT5, to put it mildly. Especially since MT4 is not yet finished.
What is "RadarScreen and many other things..."? Can you tell me more about it?
 
Will non-standard TFs be supported by MT4. For example, M90 or M45 or D2? <br / translate="no"> Tests show, that not always standard TF is the most effective.

There was a suggestion to allow users to create their own views (D2, ... XO ...).
As if that's not too hard to do...
 
<br / translate="no"> There was a suggestion to allow users to create their own views (D2, ... XO ...).
As if that's not too hard to do...


I don't understand the meaning of this phrase. Is it a statement that it's difficult to generate an arbitrary TF?
If that's what you claim, I assure you that you are wrong. And it's not difficult to create a tester, either. I wrote a variant of a more or less full-fledged tester in Delphi in three days, this with complete lack of both practical and theoretical skills in this language (theoretical gaps were filled along the way). Preliminary tests have revealed the glitches, which were immediately eliminated. Now comes the additional testing. When I can say there are no bugs, I will move on to creating optimizer strategies based not on retarded brute force (it's inefficient and totally unacceptable), but on more "advanced" ones. Plus I'll add a timeframe converter from M1 (its algorithms are stupid and primitive).

I'd rather use oars, or even an outboard motor, than wait for the wind to blow.
 
I do not understand the meaning of this phrase. Is it a statement that it's difficult to generate an arbitrary TF?

Why is it not clear?
It's written in Russian ...
There was a suggestion to allow users to create their own views (D2, ... XO ...).

I don't know how else to translate it into Russian...
I offered the developers a simple method, allowing users to create arbitrary representations of price series and to test systems on these representations. Not only representations with a constant time step (like M1, M90, ...), but also such representations as XO, bars of constant volume or area, Duke, Renko, etc. You can come up with any of your own ...

Then someone suggested another way (simpler it seems).

As if that's not too hard to do...

Again, I do not know how to translate ...
It seems to be written that it's not difficult to do,
but it's written in a slightly softer form (not as a rigid statement, but as an assumption).

I wrote a more or less full-fledged tester in Delphi in three days, this with complete lack of practical skills in this language, as well as the theory (gaps in theory were filled along the way).

Bullshit ....
Such things are not done in 3 days, especially "in the total absence of ..."
I guess 3 days is not even long enough for me to formulate my own requirements for a tester.
Maybe you just have a very superficial idea about it ...
And your tester more or less corresponds to these ideas...
Reason: