A better way to do the optimation back testing?

To add comments, please log in or register
thomas2004
337
thomas2004  

Hi all,

Quite often I will do the optimation backtesting. Since this is quite time costly expecially when you use the model of "Every tick...", I will use the model of "Open price only..". But I find if I use the optimated parameters to do the backtesting with "Every tick.." The result is quite different, much worse.

It seems what I did optimation is useless.

I've also tried to change my code and just trade not by every tick by just by bar open. But this seems doesn't help a lot.

If I change to use the model of "Control points...", it is a little bit better.

Has someone idea?


Thomas

Sergey Golubev
Moderator
106160
Sergey Golubev  
thomas2004
337
thomas2004  

Thanks.

But I use the MT4, not MT5. The links you recommend seem for MT5?

Sergey Golubev
Moderator
106160
Sergey Golubev  
thomas2004:

Thanks.

But I use the MT4, not MT5. The links you recommend seem for MT5?

Yes, it is MT5.

Besides, if you are trying to optimize/backtest EA to be very close to reality so it is backtesting with every ticks.
Just my opinion.

thomas2004
337
thomas2004  
Sergey Golubev:

Yes, it is MT5.

Besides, if you are trying to optimize/backtest EA to be very close to reality so it is backtesting with every ticks.
Just my opinion.


With every ticks? This will take days or even weeks?

Sergey Golubev
Moderator
106160
Sergey Golubev  
thomas2004:


With every ticks? This will take days or even weeks?

It depends on what you want:

  • if you want to show/to see the backtesting results which will be close to reality so it is every tick backtesting;
  • if you want to show the backtesting results to the people which will be fully identical to what they will trade - you will use MT5 backtesting with "every tick based on real ticks".

thomas2004
337
thomas2004  
Sergey Golubev:

It depends on what you want:

  • if you want to show/to see the backtesting results which will be close to reality so it is every tick backtesting;
  • if you want to show the backtesting results to the people which will be fully identical to what they will trade - you will use MT5 backtesting with "every tick based on real ticks".

My broker doesn't support MT5. So I ahve to use the MT4. If I use thw "Every ticks...", this will take days or even weeks for optimation? Assumed I backtest for 3 years and in 15M timeframe and have 3 params to be optimated.

Sergey Golubev
Moderator
106160
Sergey Golubev  
thomas2004:

My broker doesn't support MT5. So I ahve to use the MT4. If I use thw "Every ticks...", this will take days or even weeks for optimation? Assumed I backtest for 3 years and in 15M timeframe and have 3 params to be optimated.

Some coders are converting their EAs to MT5 just to test/optimize/find good settings with "every tick based on the real ticks".

There are two ways to do:

  • you are trading by your EA (on demo or live account) for at least 3 months or longer period of time (depends on your EA and timeframe), or
  • you are backtesting your EA (MT5) with "every tick based on the real ticks" which will be the same as you are trading for 1 or 2 years (exact same) with the broker. Besides, you can optimize/backtest your EA with cloud, and it will take the minutes to do.

It was about MT5.

------------------

As to MT4 so ... if you are doing it for yourself only so it is not a probem. But if you will show your results to somewhere so your backtesting results will be nothing with the reality.
It means: I am recommending to optimize/backtest with every tick mode.

------------------

Anyway - up to you. I just explained my opinion.

Marco vd Heijden
Moderator
8182
Marco vd Heijden  
You can also just skip it.
thomas2004
337
thomas2004  
Sergey Golubev:

Some coders are converting their EAs to MT5 just to test/optimize/find good settings with "every tick based on the real ticks".

There are two ways to do:

  • you are trading by your EA (on demo or live account) for at least 3 months or longer period of time (depends on your EA and timeframe), or
  • you are backtesting your EA (MT5) with "every tick based on the real ticks" which will be the same as you are trading for 1 or 2 years (exact same) with the broker. Besides, you can optimize/backtest your EA with cloud, and it will take the minutes to do.

It was about MT5.

------------------

As to MT4 so ... if you are doing it for yourself only so it is not a probem. But if you will show your results to somewhere so your backtesting results will be nothing with the reality.
It means: I am recommending to optimize/backtest with every tick mode.

------------------

Anyway - up to you. I just explained my opinion.

Many thanks!

To the historical datas, normally the historical datas from the broker are only 1-minute datas. If I backtest it with "Every ticks...", maybe it is better to buy from the third provider since they provide real tick datas such as TickStory?

Marco vd Heijden
Moderator
8182
Marco vd Heijden  

Yes buy it.

Buy all of it.

12
To add comments, please log in or register