When testing experts on the “all ticks based on real ticks” model, I found errors in the tick history downloaded from the MetaQuotes-Demo server.
EURUSD and GBPUSD broken ticks in December 2014:
Be careful when testing experts, because these errors affect the results of tests.
Similar “broken ticks” have been seen on other instruments at various times.
stringo
Please show a screenshot of the Visualizer
vanches
I think I understand the question. Although it is not directly related to the topic of the branch, but I will try to answer!
ragadop67b, you are using the MT4 tester in the “All ticks (…)” mode, so the flow of ticks will be built on the basis of OHLC data based on the smallest available timeframe.
Let’s assume that you have loaded the OHLC data of the 1 minute timeframe. At night, there can be only a few ticks inside the minute bar, and during active sessions, the number of ticks per minute can be a hundred times more. That is why the time of building bars in the tester will not be uniform, because you use the slider to adjust the speed of synthesizing ticks, and not the speed of time flow.
If you need to test based on real ticks, then you can do it in MT4, but you need to prepare the tick data yourself. Or, take advantage of MT5, but taking into account the problem voiced by the topic starter, testing for some time periods will not be correct (-:
komposter
Do you want the tester to pause between ticks like in real life?
I can’t understand any more than I am the only one, explain more clearly.
paragon67
About what uniformity of playback, and even synchronicity in real time at a speed of 5 !?
I write about the non-uniformity of the time interval between bars, let it be 1 minute or more, at any speed of the engine.
genino
The mystery lies somewhere here:
Forum on trading, automated trading systems and testing of trading strategies
Tick history MetaQuotes-Demo
paragon67, 2017.11.10 20:27
when testing the expert Advisor on a 1-minute TF, I moved the visualization engine to the value of 5
I also do not understand the essence of this post.
komposter
Do you want real-time tick testing to last as long as the history is being tested? I.e. 24 hours of testing to last 24 hours?
And why on the 5th speed, and not on the other?
paragon67
Good afternoon!
I want to support Ivan about the question “all ticks based on real ticks”.
Why does the smallest TF ,1 min, according to the tester differ in the time interval from real trading, where,when switching to each new minute bar, the tick time does not correspond to the real time interval of 1 minute (60 seconds)?
Did videosensor,but since the volume of the video file more than 300MB,made photoscreen result,where when you test the EA on a 1 minute TF, shifted the engine of visualization in a value of 5 ,increasing the scale of the grid to coincide with each minute bar,more along the lines.
I installed an online stopwatch nearby. Each measurement started with a new bar and stopped its end.
See the measurement results for the so-called “all ticks based on real ticks”, at the moment, the time when switching to a new minute bar.
As a result, the tin and how can you generally test correctly, especially to optimize the strategy on cropped ticks in time between TF bars ?
stringo
Thank you for your message. We will edit the use of ticks for history in the tester.
Ticks, although not very good quality (they can not be called broken), but they can be correctly applied to the story. Will do
vanches
the first image shows the visual testing window for the all ticks model:
the second image shows the visual testing window for the all ticks model based on real ticks:
In the screenshots, we see the same tool at the same time. Find 10 differences!?
By the word “broken” – I mean corrupted ticks from which not correct bars are built, and therefore not correct test results are obtained.
sergeev
what are you trying to show in the screenshots? konretno show where there is incorrect ticks. What does “broken” mean ?