Hi rich people,
after automatic update from build 920 to 924, the same backtest ended with a different result.
Since build 924, the spread setting is no more taken into account when backtesting using a read-only history FXT file (and only when the FXTfile is read-only)
Using build 920, spread setting is correctly used for backtest, and log shows "Test generator : spread set to xxx" (see screenshot). Good and working fine.
After build 924 update, the "Test generator : spread set to xxx" is not seen anymore in log and backtest is false (not computed with the correct spread)
Why using a read-only history file ?
Because I am doing parallel backtesting using virtualisation, with a multi-core host server and several single core VM.
History FXT file needs to be read-only and defragmented on the VMs, else IO delays are raising to 20-30%, thus slowing down all the backtests.
Have you reported this bug to the service desk?
Same here for me under MT4 build 950.
Using the 'Every Tick' model in strategy tester doesn't use the Spread set anymore.
Instead the Spread is now 0 by default.
To work around this bug:
1) Switch the test model to 'Open prices'
2) Now select or input a new Spread.
3) Start and stop backtest.
4) Change the model back to 'Every Tick'.
If you now start the backtest you will have the new Spread set also in this model.
Why should anyone report anything beyond informing other users ?MT4 is not an open source product - it is a commercial effort, and as such should not reply on public testing.
Why there is no QA/developer monitoring these forums ?If you MT4 wishes for users to test its products it should offer suitable compensation or open-source it.The problem is a long standing absence of any real software development process behind this product which renders any production use beyond 10 lines of code retail public suicidal.
YES, my post here is the exact copy of the Bug Report I've sent to MetaQuotes (at the same time)
This copy here is intended to inform other users like me.
I wastes hours, first to understand that backtests were false, and also to understand where is the source of the problem.
So my goal is first to let other people know and explain why their Backtest could be wrong, and also save other people search time by pointing where the problem is.
Thanks for your workaround, this is why sharing between users is usefull.
For now I've blocked auto update in order to run build 920.
No problem, but it's been already nearly 3 months if I can see it properly.
They usually answer within a month or never.
To add comments, please log in or register