Analysis of test results and optimisation in the MetaTrader 5 strategy tester - page 15

New comment
 
Ilya Malev #:
Not everything is so simple.
But the most despicable is when the tester considers the maximum Ask missed in the "Open Price Only" / OHLC / All ticks mode as a stopout trigger. All ticks are modelled based on the data of minute bars, and not on the Real ticks, the main consequence of which is the LACK of information about the spread inside the minute, i.e. the spread is selected by the algorithm the same for all modelled intra-minute ticks (as in the case of other TF in the modes "Opening prices only") for the whole bar. As a result, even if we refuse all types of orders, including SL and TP, and orient the algorithm only on Bid prices, the Real ticks mode will screw up part of the tests, while all other modes will pass normally.
 

Ilya Malev #:
"Аттракцион щедрости" в случае проскальзываний лимитов к ценам ближайших тиков

Stats.

on potic models like "OHLC" does not cancel it.

A generosity attraction for stop levels. I optimise through Virtual myself, so I can turn sliding on/off. However, sliding for stop levels is always on. Grails are not needed.

 
Ilya Malev #:

It's not that simple. Imagine a sequence of real ticks inside some bar

...

Show 4 control ticks that must be selected in this bar for all these orders on this bar to be triggered, as it will be in the "Real ticks" mode.

The principle of ascii generation is unfortunately not disclosed on the MQ side. I see that they do not coincide with the ones that were in reality, so there will obviously not be a full correspondence with real ticks, and the triggering of orders - too.

I think, with such (postictic) requests to the quality of testing, we should limit ourselves to OHLC M1 mode or generate a custom symbol with thinned real ticks according to our own principles (there are examples). It will be already very fast in any case, and chasing the mode at opening prices clearly implies a willingness to be more loyal to accuracy.

 
Stanislav Korotky #:
I think, with such (postictic) demands to the quality of testing, one should limit oneself to OHLC M1 mode or generate a custom symbol with thinned real ticks according to one's own principles (there are examples). It will be already very fast in any case, and chasing the mode at opening prices clearly implies a willingness to be more loyal to accuracy.

That's what I went down the path of creating a custom symbol in the beginning. I thought that turning OHLC into OHHLLC (O-Hbid-Hask-Lbid-Lask-C) would solve the problem. But it didn't solve the problem - it only introduced more discrepancies in the tests, so I didn't bother to look into the reasons further. Now I understand what the reason is - in the different mechanism of triggering orders in the Open Price Only mode and the Real Tick mode (for OHHLLC to work, you need to run the custom symbol in the Real Tick mode). Because in the first case the triggering will go without sliding, and in the second case with sliding, and if you make six ticks from one bar, especially if the bar is not M1, but for example H1, this sliding will make the test results obviously inadequate.

And take into account that the All ticks mode generates a stream of ticks from minute bars, i.e. within a minute the spread is fixed by the value that is set once in a minute bar (most likely by the value that was at the opening). Real ticks are the only adequate intra-minutes model in MT5 tester. I did not realise this for a long time and used All ticks everywhere in tests. Until one day recently I got a picture of a completely opposite profitability chart with the All ticks and Real ticks modes on one of my Expert Advisors (in that Expert Advisor there is a very close to the price trawl of orders, which was made within the spread).

 
Ilya Malev #:
sliding will make the test results obviously inadequate.
I assume that the trading mode of the account on which the tester is running or certain specifications of the symbol can enable the mode in the tester, when there is no sliding on real ticks.
 
fxsaber #:
I assume that the trading mode of the account on which the tester is running or certain symbol specifications may enable the mode in the tester when there is no slip on real ticks.
Please clarify which setting you are referring to
 
Ilya Malev #:
Please specify which setting you are referring to
You have to look up the documentation. Netting, flippers, exchange, etc. For example, on an exchange, a limit must be executed exactly at the price. Most likely, this is taken into account in the Tester.
 
fxsaber #:
We need to get the documentation up. Netting, flippers, exchange, etc. For example, on the exchange, the limit must be executed exactly at the price. Most likely, this is taken into account in Tester.
Anyway, you can't edit the account mode even in a custom symbol. I've noticed that even setting my own margin call and stopout level in the tester doesn't work, maybe it was set in the "account mode settings" of a particular terminal/broker that it can't be changed in the tester, when I checked, I don't know :).
 
Ilya Malev #:
Anyway, the account mode cannot be edited even in the custom symbol.
Find the required demo account and run the Tester by connecting to it.
 
fxsaber #:
Find the demo account you need and run the Tester by connecting to it.
That's how I ended up doing it.
1...8910111213141516171819202122
New comment