Advice on not using the MetaTrader 4 Strategy Tester - page 4

New comment
 
>>>excessive clarification for someone who has been interested in Forensics to varying degrees for over 10 years :)

!!!!!!!!!
 
Renat писал (а):
All bars are based on bids.
It turns out that incorrect statements are made from an empty place due to ignorance of the existence of Ask price.

The first two examples are solved. I'm waiting for clarifications on the third.

P.S. Renat, I beg you very much not to make hasty conclusions. We can all see that you have a hot temper. But there are certain limits. I think you also don't know a lot of things in this world, especially in areas that are not directly related to your work. There's no shame in that. Special thanks for explaining to me the technology of bar building. Moreover, I`ve read about it before and it`s just not imprinted in my mind. Now I will remember it.
 
solandr писал (а):
>>>explanatory for someone who has been interested in forex to varying degrees for over 10 years :)

!!!!!!!!!

What would that mean :)

The difference between Ask and Bid, as well as the terminology, was something I learned for the first time in 1998 (I don't even remember exactly), when I tried to go from theory to practice and opened my first demo account with fxeuroclub. I would immediately understand everything as soon as I tried to do something, but I've never thought about the method of drawing bars until now :) I did not think about it at all. I prefer figures. A chart is a subjective thing.
 

A snippet from recent history.

EURUSD M30 - tester:



Position opening condition : 

// по расширению канала Дончиана больше первого диапазона опорного центра
 
   if(I2Current-I1Current>R1-S1        // текущий канал Дончиана шире канала опорного центра
      && I2Previous-I1Previous<R1-S1   // предыдущий был уже
      && I2Current>I2Previous          // верхняя граница канала Дончиана пошла вверх
      && I1Current>=I1Previous         // нижняя - не пошла вниз
      && MP>0                          // локальный тренд +
      && Ask<High[0]-7*Point           // покупка не на максимуме цены    
     )
      signal1=1;

As can be seen from the chart, the condition worked in the tester. Not in real time on the demo. So far I have only one explanation: price has not bounced from the maximum by enough pips in real time (last line). The other lines were executing. MP variable too (I have a separate indicator). It is calculated like this:

   MP=((Close[0]+Close[1]+Close[2])/3
       - (Close[0]+Close[1]+Close[2]
          +Close[3]+Close[4]+Close[5]
          +Close[6]+Close[7]+Close[8])/9)*1000;

Can anyone confirm or refute my suspicions that in this case the data simulation results in the tester did not match the real price series?

 
>the condition on the tester worked. In real time on the demo it didn't.

This can be explained in the following trivial way. A requote happened on the real market and you had a small slippage in the OrderSend function (many beginners usually set slippage for opening at market equal to 0!!!), or the Expert Advisor just did not handle this situation with non-opening of an order. I always set slippage for myself equal to 5 pips when opening at market. As a result, I have not yet noticed any problems with opening orders, although I am constantly comparing the tester results with real ones.
 
2 Rebus
Real price series and they don't have to match the results of their modelling in the tester. They don't owe anyone anything at all.
This thread started with a comparison of simulation results on different timeframes. And you jumped to comparing real and simulated series.
 
Interesting fact: while testing with 197 version of MT we get one data at 90% quality, while testing with 193 version and getting history in real time we get different data at 90% quality too. So where is the truth? Of course, the differences are not drastic, but they are significant.
 
2SNSH
Try version 157, if you can find it with someone. Probably that's where the truth is buried deep down! As they say, it's a ruddy old horse, isn't it? :o))))))))))
 
solandr писал (а):
2SNSH
Try version 157, if you can find it with someone. Probably that's where the truth is buried deep down! As they say, it's a ruddy old horse, isn't it? :o))))))))))

I meant is there any difference when downloading the history from the Date bank and getting it in real time over the same period from the same DC and the impact on testing.
 
solandr писал (а):
>the condition on the tester worked. In real time on the demo it didn't.

This can be explained in the following trivial way. A requote took place on the real account and the slippage in the OrderSend function was small (many beginners usually set the slippage at market opening equal to 0!!!), or the Expert Advisor just did not handle the situation with non-opening of an order. I personally always have slippage equal to 5 pips when opening by market. As a result, I have not noticed any problems with order opening yet, although I am constantly comparing the tester's and real-time results.
I have 3. I will try to set it to 5. Actually, it is clear that modeling is modeling. Apparently, the problem will be solved only when MT4 will store and use a tick chart for testing. Even if it results in huge volumes - it is better than looking for a black cat in a dark room.

Honestly, it's a terrible shame that I forgot the basics. I mean the bar building. It's been too long of a break. Spent almost four years building a house, so I've almost forgotten everything else. Only about half a year ago I started to get interested in the market again. Like from scratch, honestly :). Although, of course, I'm lying. Things are simpler than then. And with autotrading there was even a purpose in life. (Wouldn't turn out to be just a skyline :).
12345678
New comment