Open project - tester-optimiser in-house - page 6

New comment
 
Well, if you formulate the requirements for a tester over three days, then yes.
I can only feel sorry for you.
If it's not a secret, could you tell me what the requirements are for a tester? Or maybe I'm flying in the clouds and you're taking me down...
 
If it's not a secret, could you share the requirements for the tester? Or maybe I'm flying in the clouds and you're taking me down to earth...

I won't ... :)
I don't have plans to make my own tester yet ...
 
Yeah, there's more to it than that.
Dart - if you're such a mega cool programmer who figured out Delphi in a couple of minutes.
Help Rosh write a tester for MQL4 and I'm sure with your speed you'll be up and running in seconds.
 
I'm not programming on McLean anymore. I'm sick of it. And the language sucks. And I'm not a mega-cool programmer. It's just simple things everyone should do.
But I can formulate the requirements for tester, maybe you can criticize them and I will make corrections in my tester.
So, the requirements:
1. tester must emulate incremental time, so that you do not bother with it inside the strategy.
2 Tester should provide procedures for working with orders (installation, modification, and removal)
3 Tester must be on the occurrence of conditions:
a) Close orders on the stop (always the stop has priority)
b) Close orders on profit
c) Open pending orders.
d) When the chart is finished, the tester should delete all orders
(I'll tell you a terrible secret, he does all this using the standard procedures)
The tester should not look inside the bar, except for cases when it closes/opens on pending orders or on stop/profit.
This is the basic one. There are also statistics calculations, order management procedures, data conversion from *yu.csv to arrays, function-methods. This is what we have.
But it will be steeper. I have caught the main bugs, and I am too lazy to test the tester further. As soon as the testers get feedback, I'll continue.
Ah, consider it my help.
By the way, I'm not the only one who wrote my tester. BiT from Viac doesn't trust black boxes either.

Also. A tester should be written for something, not for the sake of the tester itself. Don't you have enough of an MT3 tester?) I am writing it with the purpose of adding there algorithms of strategy optimization, that's why I have special requirements for speed. And what they show you here
[quite]
VC++: Time: 47 ms, ResInt=-752584127 ResDouble=17247836076609
MQL4: Time: 797 ms, ResInt=-752584127 ResDouble=17247836076609
MQL2 : Time: 22156 ms, ResInt=333333383333717340000 ResDouble=3333333833717340000
WL3 : Time: 32422 ms, ResInt=3.3333333383333717E20 ResDouble=3.33333333833717E20
[/quote]
I don't know, I haven't checked it myself, but in my programmer book it says, that conversion from timer in Windows98 is generated every 55ms, in XP every 10ms, maybe author is a fool, but where does 47ms, 797ms come from?
 
Well, I guess we'll have to wait for the tester from the developers after all...
 
[qoute]
Ok, I guess we'll have to wait for the tester from the developers after all...
[/quote]

To each his own.
 
I don't know, I haven't checked it myself, but in my programmer's book it says, that conversion from timer in Windows98 is generated every 55ms, in XP every 10ms, maybe author is a fool, but where does 47ms, 797ms come from?

Since you took up the criticism, you should check operation of standard WinAPI function GetTickCount() (not on Win98). Otherwise you may get:
1) based on theoretical reasoning - "it's easy to write a tester".
2) by "my programming books" - "the author is a fool".
 
<br / translate="no"> 1) based on theoretical reasoning - "it's easy to write a tester"
2) based on "my programming books" - "the author is a fool"

Heh.
And what is difficult in writing a tester? If not a secret or comtayna. Or maybe I stated something wrong there? I only listed what I did myself. Maybe I missed something?
I've never claimed to be the absolute truth.
Is there any way to get increments not from 1ms but from 1μs or lower?
 
Haven't been here in a while. I'd forgotten that Dart was here - he seems to be very allergic to MT :)

I looked, 15-20 people have already downloaded what I posted (I can't call it a tester, rather a mock-up). I don't know what will be the tester in MT4, but it will not fit all occasions, there will always be some specific task, which it does not fit. Therefore, do not forget about the possibility of creating your own tester-indicator.
And it can be used for:
1. portfolio strategy optimisation (said to be out of the question)
2. genetic optimization methods
3. Gradient-based optimization methods using momentum
4. Neural network training

This is what I just remembered. And there may be at least as many other tasks.

PS. Additionally. tester's functions could be redefined in dll (in C or other languages). Although I haven't studied this question yet. This will give greater productivity and debugging capability.
 
I've been fiddling with the tester at certain points too.
Some conclusions:
1) For proper testing the
Minutes for M15 and lower (ticks are an excessive luxury)
5-minutes for H1 and M30
15 minutes for H4, etc.

2) What to do with swaps?
3) To test crosses you need to be able to convert their currencies
To USD for each moment in the past. I.e. in addition to GBPCHF
USDCHF and GBPUSD quotes array.
1234567
New comment