Spread in backtesting.

[[Deleted]]

How does the backtester get/use the spread in backtesting?
The reason I ask this question is that I get different backtest results depending on when I run the backtest (same data, same computer, same server etc).
I can only assume the 'live' spread is been used, hence the variation in results.

[Renato Fiche Junior]

Craig,

At "Strategy Tester" section (press F6 key), click on "Symbol properties" button to see the spread used by the tester.

[[Deleted]]

Thanks for the reply, where do these settings come from? or are they just hardcoded?

[Ralph Ronnquist]

Craig:
Thanks for the reply, where do these settings come from? or are they just hardcoded?

I wouldn't know for sure, but from stringo's article 'Testing of Expert Advisors in the MetaTrader 4 Client Terminal: An Outward Glance', it looks like a "current spread" is obtained from the server when the test cache (the FXT file) is created, and then is re-used on re-run. Running at different times and different computers or re-calcualting the FXT file give ample opportunties to cache different spreads (if it changes).

Another question is whether the backtesting also models the "overrunning" behaviour of a real run, i.e. "dropping" ticks for a busy EA. If that's the case, then you would have a new dimension to the possibilities of differences between platforms and runs.

[[Deleted]]

Hmmm, once again, thanks for the answer.
Does anybody get the feeling this could all be much easier?

Another question, if I load up M1 data in the history center, what is the difference between using the 'period_convertor' script and
letting the backtester recalculate a higher timeframe on every tick? do they not effectivly do the same thing?

[Slava]

See 'FXTHeader'

May be it helps

[[Deleted]]

Not sure, does this mean there is no simple answer?

[Slava]

Spread value is stored in the FXT header as 4-bytes integer with the offset 124 bytes. This is a constant and cannot be changed during the entire testing. You can change it in the header yourself. But it remains a constant and will be applied to the entire testing.

Answering your other question: There is a difference, yes. The ticks are generated inside of M1. Of course this is not simple. Please read article 'Strategy Tester: Modes of Modeling during Testing'

[[Deleted]]

Cheers, I'll check it out.

[[Deleted]]

Um... that clears it up a bit.
As a bit of feedback, I know English is not the first language of the people involved in metaquotes, and I appricate that the effort has been made to translate the text.
But some of the articles are nearly unreadable ("My first grail" for example), it's almost as if the text has been run though google translate.

[Maria Sountsova]

Craig:
Um... that clears it up a bit.
As a bit of feedback, I know English is not the first language of the people involved in metaquotes, and I appricate that the effort has been made to translate the text.
But some of the articles are nearly unreadable ("My first grail" for example), it's almost as if the text has been run though google translate.

Hello Craig,

I translated articles into English. You wrote "some of the articles are nearly unreadable". Does it mean that the others are ok? :)
Then it is not the matter of translation, but of the author's vision, isn't it?
I would appreciate very much if you could contact me (to discuss these matters) via ICQ (92124862) or msn ([e-mail]).
Anyway, thank you very much for your feedback.

Best regards