Development plans for the MetaTrader 5 Strategy Tester - page 14

 

Tried to do Optimisation by real ticks without filtering. For this I had to disable RAM-Drive and work with the tester via SSD.

SSD blinks all the time during optimization. Some wild activity from the Tester's side. This despite the fact that it takes 30 seconds for each pass.

These Agent\temp\bar*.tmp files are several gigabytes in size for what? Why read them all the time during Optimization?

 

What are these files? Tester\cache\*.tst

 
A start/stop button visible on all tester tabs. When debugging the Expert Advisor through the log, there are a lot of unnecessary actions. Press start, go to the logs tab, look at it, go to the settings tab, press start, etc.
 
npats2007:
Start/stop button visible on all tester tabs.

The general one.

 
fxsaber:

Is it possible to keep only one copy of the price data for all local Agents in RAM? Right now, memory consumption is quite inefficient.

This all has to be redesigned, now the agents are independent Windows services. I'm asking at least to learn how to distribute files to one PC, and then copy them internally for agents, and it's not working yet.

In general, of course, it seems reasonable, but probably will slow down the work of agents, because it's not possible for each agent to read the same memory address cell synchronously, and there will be a queue... or am I wrong and technology has gone too far?

 
Renat Fatkhullin:
  1. We will drastically expand the detailed testing settings with commissions, order triggering conditions, slippage, etc.

The order triggering conditions are very lacking. On the Moscow Exchange, how to test strategies on limits in the current tester is not clear. It would be nice to add partial limit order execution depending on volumes.

 
fxsaber:

It feels like this problem is only on MQ-Demo accounts. I haven't noticed the bug on the other servers.

It is partly possible to work if you drag a symbol from Market Watch into the Tester with the mouse. Wasn't aware of this feature before.

But still there are custom symbols, which even this way you can't drag into the tester.

 
fxsaber:

What are these files? Tester\cache\*.tst

Single test cache.

Also added ability to save optimization cache by symbols from market review. And format of optimization cache header has changed

 
Slava:

the format of the optimisation cache header file has changed

I will wait.

 

Here are the earlier thoughts on improving the MT 5 tester.

Forum on trading, automated trading systems and strategy testing

MetaTrader 5 platform beta build 2055: Integration with Python and massive improvements in strategy tester

Askr, 2019.05.17 06:34

No new features and bug and bug fixes have been made in the tester?

Specifically.

1. has the bug with testing on mosbirge stocks been fixed? What caused the false counts on real ticks?

https://www.mql5.com/ru/forum/307943

2) Is the tester adapted to foreign markets? The CME? With its intraday and daily margin?

https://www.mql5.com/ru/forum/309332

3.

In the mt4 tester you can stop the test at any time and see all trades and the overall result of the test, while in mt5 you will only see a white sheet, that is nothing.

When you stop your test, you may see only a white sheet, and there is no point in waiting until the end of the test.

https://www.mql5.com/ru/forum/189085/page21#comment_10910219

4.

We miss the complete analysis showing results for each month being tested, like in monitoring services.

For example January 19- percentage

February 19 percent

March 19

It would be better to show the drawdown of course, although it is not the main thing.

But it would be better to show the monthly yield because it's impossible to show the yield for one month if the graph is straight and for a long period of testing.

https://www.mql5.com/ru/forum/189085/page21#comment_11488482

5.

It's a trifle, but very inconvenient that the tester has no choice ofhistory of positions, but not deals.

It is much easier and faster to compare trades between the tester and a real account.

https://www.mql5.com/ru/forum/189085/page21#comment_11488482

6.

The average size of profitable and unprofitable trades in pips is not enough.

https://www.mql5.com/ru/forum/189085/page22#comment_11573743

7. And overall suggestion, add a bit more features that are in the same account monitors.

8. Limit the number of cloud agents/PRs involved.

https://www.mql5.com/ru/forum/298902

Reason: