Optimizing Strategy

New comment
 

To optimize a MA cross strategy, is it necessary to use tick data for every step of the process to get the most accurate results or could testing with tick data be reserved for the final step in the process? For example, let's say the MA period is being optimized over a 6 year time span and we start off testing periods by steps of 5 from 5 to 65 using open price modeling. Let's say the optimized result of the testing results in a period of 20 and we now refine the process by testing by 1 step from 15 to 25 using every tick modeling. Will using open price modeling for the first step of the process have made a significant difference for its result instead of simply using every tick modeling for both steps of the process?

My goal, if feasible, is to delay or avoid the excessive use of memory that accompanies the utilization of every tick modeling.

 

To optimize a MA cross strategy, is it necessary to use tick data for every step of the process to get the most accurate results or could testing with tick data be reserved for the final step in the process?

That depends on how it's programmed.

Example1: if(ma_red_current_value > ma_blue_current_value){..do something..}

Example2: if(ask>horizontal_line_value_x){...do something....}

In the above examples, it will matter if you're using tick to optimize or not. However below is an example where it wouldn't matter as much.

Example1:

if(TimeCurrent()==iTime(Symbol(),0,0)){

if(ma_red_current_value > ma_blue_current_value){..do something..}

}

Will using open price modeling for the first step of the process have made a significant difference for its result instead of simply using every tick modeling for both steps of the process?

Program it for open price then use open price to optimize. Program it for ticks then use ticks to optimize. {Remember: unless you customized mt4 to use real-tick data, your ticks in the backtester is 1_Minute OHLC}. When it comes to optimization, there are 2 things you wanna keep in mind. 1) Over Optimization or Curve-Fitting. 2) Recent Optimized results, holds more weight than older values.

 
ubzen:

To optimize a MA cross strategy, is it necessary to use tick data for every step of the process to get the most accurate results or could testing with tick data be reserved for the final step in the process?

That depends on how it's programmed.

Example1: if(ma_red_current_value > ma_blue_current_value){..do something..}

Example2: if(ask>horizontal_line_value_x){...do something....}

In the above examples, it will matter if you're using tick to optimize or not. However below is an example where it wouldn't matter as much.

Example1:

if(TimeCurrent()==iTime(Symbol(),0,0)){

if(ma_red_current_value > ma_blue_current_value){..do something..}

}

Will using open price modeling for the first step of the process have made a significant difference for its result instead of simply using every tick modeling for both steps of the process?

Program it for open price then use open price to optimize. Program it for ticks then use ticks to optimize. {Remember: unless you customized mt4 to use real-tick data, your ticks in the backtester is 1_Minute OHLC}. When it comes to optimization, there are 2 things you wanna keep in mind. 1) Over Optimization or Curve-Fitting. 2) Recent Optimized results, holds more weight than older values.


Thank you for your reply. This has helped me out a lot.
New comment