Errors, bugs, questions - page 2384

 
Slava:

It's been three years since the maximum spread was fixed, not the minimum. It seems that they haven't corrected it in the help

Does the quote provider (broker) do this or does the server do it automatically and a particular broker cannot do it differently?

 
Ilya Malev:

Does the quotation provider (broker) do it or does the server do it automatically and a particular broker cannot do it differently?

The server does it automatically
 
I don't know what your reasoning was three years ago.
Why it is wrong: It is an axiom in any testing of any Expert Advisor: if it is possible to test the most and the least favorable scenario in each specific case, you should take the least favorable one. This rule has no exceptions, unless the trader wants to look at his/her trading systems through rose-colored glasses, and thus lose money.
If we set the minimum spread per bar, as we do now, limit orders are "falsely" triggered based on more "favorable" spreads than they really were, including takeprofits, and stop-losses do not trigger where they should have. In general, almost any transaction is performed at a better price than it should be in reality. Is this what you wanted when you changed the correct order to the current one?
 
Ilya Malev:
I don't know what your reasoning was three years ago.
Why it is wrong: It is an axiom in any testing of any Expert Advisor: if it is possible to test the most and the least favorable scenario in each specific case, you should take the least favorable one. This rule has no exceptions, unless the trader wants to look at his/her trading systems through rose-colored glasses, and thus lose money.
If we set the minimum spread per bar, as we do now, limit orders are "falsely" triggered based on more "favorable" spreads than they really were, including takeprofits, and stop-losses do not trigger where they should have. In general, almost any transaction is performed at a better price than it should be in reality. Is this what you wanted when you changed the correct order to the current one?
Test on real ticks
 
Slava:
I will use real ticks to test it.

Thank you, but you have a good idea of what optimization of thousands of variants of multicurrency strategy tests on real ticks means, don't you? Of course I will not follow your advice, but I will follow fxsaber's one

That is, I will do manually, spending a lot of effort, what you should have done, even based on the manual to the terminal, and not only on common sense, but for some reason you did the opposite, and you don't even want to explain why
 
Slava:
Test on real ticks
Your answer could be understood as you are deliberately screwing up the spread so that everyone is tested on real ticks.
 
Slava:
Test on real ticks

The cloud for real ticks and custom bars is not available. Really don't use it.

 
Ilya Malev:

That is, I will do manually, spending a lot of effort to do what you should have done, even based on the terminal manual. not just common sense, but you did the opposite for some reason, and don't even want to explain why

You don't have to do it manually. It is enough to automate once and completely forget about the problem.


And you don't resent the fact that the marks are marquets and therefore slip. Or that limit orders on a Hedge and a non-exchange symbol slip to the plus side, giving a huge imaginary increase in the result.

 
fxsaber:

You don't have to do it manually. It is enough to automate once and completely forget about this problem.

Every time before the test on new dates will be necessary to generate from ticks partially or completely by a script the minute history of each traded instrument with the "correct" spreads, or can it be done in a simpler way?

Besides, symbols with different names will be used in trading and in tests, which additionally creates unnecessary confusion.

 
Ilya Malev:

2. Regarding custom characters, can the terminal do this LowBid to LowAsk replacement automatically, or do you have to build it yourself by calling CopyTicksRange and CustomRatesReplace?

The substitution will not work, because LowAsk > HighBid happens. That's where the spread should be calculated.

If you don't look at Cloud's limitation, then there is almost no sense to test by custom bars at all, because there are custom ticks.