Did strategy tester changed?

 

Using the same code and the exact same parameters on the same server, strategy tester is yielding very different results on different builds of MT5. Did something change or is it just me? It started for me after Build 3490.

Attached are two images of backtest chart from the two different builds:

MT5 build 3446 vs MT5 build 3492

Another change I noticed on the latest build is the strategy tester finish running very fast, like 10x faster or more than on the older build. As the result is very different between the new and old build, I need to know which is the reliable one.

The Fundamentals of Testing in MetaTrader 5
The Fundamentals of Testing in MetaTrader 5
  • www.mql5.com
What are the differences between the three modes of testing in MetaTrader 5, and what should be particularly looked for? How does the testing of an EA, trading simultaneously on multiple instruments, take place? When and how are the indicator values calculated during testing, and how are the events handled? How to synchronize the bars from different instruments during testing in an "open prices only" mode? This article aims to provide answers to these and many other questions.
 

You are not alone, my friend. I have been having the exact same issues since a recent update.


While I haven't found a solution yet, I wanted to add some additional details.

Problems are arising with "Every tick based on real ticks" selected. To test this, I wrote a simple EA which prints the date and time on every tick. Using real ticks, there are far too few print statements, and they only occur at the start and end of a bar.

This is happening for the same symbols I used several days ago that do have full tick data available. I also tried creating a new symbol and importing full bar and tick data but this did not change the issue.

Given that such few data points are being processed, it is no surprise that we are seeing so much runtime speedup!


I also want to note that the other modelling options such as "Every tick" and "1 minute OHLC" seem to work as expected. 

Hopefully someone can figure this out soon as this issue has put a halt to my testing! I have also not been able to figure out how to revert to a previous build.

 

Struggling with the same issue today. As fxsaber indicates in the other thread in the russian subforum, real ticks are computed as expected when selecting M1 in the strategy tester timeframe. However, if you use a multi-symbol EA, only the one selected in the strategy tester will be using real ticks, even using indicators in the EA explicitely in the M1 timeframe for all symbols.

I haven't found a workaround this, so for now all EAs which have to be tested with real ticks won't work... I hope it get's fixed as it seems like a unintended bug, seeing that simulated ticks work properly.

 
Good news! Seems like the issue has been fixed in the last update, my real tick based EAs apparently show the same result as they previously did.
 
J_Guerrero #: Good news! Seems like the issue has been fixed in the last update, my real tick based EAs apparently show the same result as they previously did.
By latest update do you mean build 3500?
 
Fernando Carreiro #:
By latest update do you mean build 3500?
While I can't speak for J_Guerrero, I can say that after updating to build 3500 everything is working properly for me again!
 
Fernando Carreiro #:
By latest update do you mean build 3500?
Yes, that's correct.