Inconsistent test results

New comment
 

Hi,

I've been system trading futures using a daily system (TradingBlox) for a few years - and it is very reliable and consistent.

By contrast, I've just finalised a system for forex in MQL4 - & I find that each time I run a back test on 2 years of offline historic H1 data, I can get wildly different results, even if all tests are at Tick level, and all parameters remain unchanged!

All it takes to change a winning system into a losing one is to check/uncheck the 'Visual Mode' box.

This should not change the test results, but it does - how can I know which is the correct back test ?

Also, if I run an optimisation using parameters, then chose one set of parameters to run individually, again the results differ !!

I thought I'd got to the bottom of it & that it had been a mistake on my part - so I went live & am trading a $25k account. However when testing against other currencies I find that it still produces different results without any changes to the code or parameters. What is going on here ?

Anyone else had this experience ?? Is this platform reliable enough ??

Thanks, Dave

 

Take terminal offline when you are testing. When you click to start your test, the terminal will look for the latest known spread from the server (among other things) and apply that across all back history. So if you are online with a variable spread broker, you will get different results as spread varies.

If you search, you will find some other solutions including a code to manage the spread...I don't have the link handy.

hth

V

 

Thanks Viffer

That resonates with what I'm experiencing - terrible test results for a few days, then good ones !!!

Now you pointed me to the spreads, I found the following posts...

I'll message the guy who said he had a solution & see if it posted anywhere.

Dave

New comment
Reason: