Backtest Result Feedback

 

Hello,

I had a theory on a winning strategy so I went ahead and created an EA to implement it. This is my first time doing backtesting or using the Strategy Tester.


Was hoping to get some feedback if these numbers are considered good and what could use improvement to be viable.


Here is the screenshot of my backtest results from Feb 1 2019 thru Nov 15 2023.


 

What testing mode did you use ?

Can you show the "Graph" ?

 
Alain Verleyen #:

What testing mode did you use ?

Can you show the "Graph" ?

I used "Expert Advisor singe test" mode

Here is the graph, numbers are smaller because I lowered position size from last screenshot.

 
itjunkiifx #:
I used "Expert Advisor singe test" mode

I meant testing modelling, real ticks ? 1M OHLC, or ... ?

 
Alain Verleyen #:

I meant testing modelling, real ticks ? 1M OHLC, or ... ?

Got it, this was ran with "Every tick" for modeling, 1ms delay, emulating slippage
 
itjunkiifx #:
Got it, this was ran with "Every tick" for modeling, 1ms delay, emulating slippage

You need to check with real ticks.

"Every tick" is suitable during the development of an EA, to check that all is working as expected but not to get reliable results.

Depending of the strategy the results using real ticks versus "Every Tick" can be similar or significantly different.

 
Alain Verleyen #:

You need to check with real ticks.

"Every tick" is suitable during the development of an EA, to check that all is working as expected but not to get reliable results.

Depending of the strategy the results using real ticks versus "Every Tick" can be similar or significantly different.

Thanks for the guidance Alain, ran from feb 2019 on real ticks and it definitely dropped my stats. Still ended up winning every month, here are the new numbers. In your opinion do you think this strategy may have some legs?

Files:
 
itjunkiifx #:Thanks for the guidance Alain, ran from feb 2019 on real ticks and it definitely dropped my stats. Still ended up winning every month, here are the new numbers. In your opinion do you think this strategy may have some legs?

The quality of your test is no good. When testing with real tick data, don't test any further back that the available tick data, otherwise it will skew your results, usually making it look better that it actually is.

If your tick data only goes as far back as 2021, then don't test from 2019. You can see this from from your "History Quality" which is only 60%

 
Fernando Carreiro #:

The quality of your test is no good. When testing with real tick data, don't test any further back that the available tick data, otherwise it will skew your results, usually making it look better that it actually is.

If your tick data only goes as far back as 2021, then don't test from 2019. You can see this from from your "History Quality" which is only 60%

Thanks Fernando, I will do more research in to the docs about that. I re-ran from Jan 2021 and that put my quality at 100%, here are new results.