• Registered: 2010-11-18
  • Posts: 1,179

Topic: Create a "Built In" optimizer EA

Hi Popov,

Is it possible to create a "Built in Optimizer" EA?

What do I mean?
1.  The EA has an additional code (besides our usual input) to function as an optimizer to change the smoothing, base and signal inputs for all the opening/closing conditions

2.  The optimization criteria/calculation is based on finding best optimal combinations (of smoothing, base, signal period) that "matches" closest to the highest peak and the lowest trough, 1 possible "indicator" to be used or to be considered, can be "previous high and previous low".

Of course, we can also consider optimize the period input to find the best combinations among all the opening/closing conditions to match closest to the highest peak and lowest trough of eg. H4, H1 etc.

3.  We can choose the period for the optimization, i.e. which period of "previous high and previous low" we want to focus or optimize.  Hence, there is an input or drop down menu to choose which time period we want to optimize.  (I wish to have 2 time period, i.e. I want to track the previous high and low of H4 but at the same time, take into consideration of 5min or 15min of "previous high and low" to keep the optimization "relevant", so that the optimization process is not only looking into past history, eg H4 but is also incorporate the current market volatility.


If such "built in" optimizer EA is not possible.  Would you consider having this method of using previous high and low as a method of optimizing our EA.  If we can identify a point of entry (based on previous high and previous low), it would be "easier" for FSB to narrow it's focus and churn out EA that has higher statistical significance based on this concept of deduction (working backward, starting with an end point that served as our EA entry point (based on previous high and low) and working to find the best combination of opening/closing conditions to match it's entry/exit point. 

Currently, we are working forward, i.e. we input our opening/closing point and see how much profit we can get.  Why don't we work, using deduction method...I already know the highest peak and trough (i.e. my entry and exit point), now using FSB to deduce/identity which indicators and opening/closing conditions matches these entry and exit point.


Better still, if FSB can identity all the previous highs and lows for a certain time period eg. H4 over a period of historical data, to come out with the best combination that matches as close as possible to all these previous highs and lows of H4 over a period of 3-5 years, etc., that would be ideal too.

Food for thoughts

  • Registered: 2016-02-14
  • Posts: 160

Re: Create a "Built In" optimizer EA

i get the feeling that the generator already optimises

too much optimisation is self defeating and makes the strategy less robust

that said i do not understand how to use the optimser is there any video available?

  • Registered: 2010-11-18
  • Posts: 1,179

Re: Create a "Built In" optimizer EA

It is HOW you optimize, not  HOW MANY TIMES you optimize.

The focus is on the rules/methods employed in the process of optimization that will increase the "quality " of EA having higher statistical significance.

As I've mentioned about using some "filtering" or trading rules to increase the Generator results, cos sometimes I get "theoretically" unsound EA.  Here I'm suggesting using previous highs and lows as one possible criteria/filtering rules/method to search for the entry and exit points/logical conditions. Instead of searching "randomly" for opening conditions (entry point) and closing conditions (closing point), using previous highs and lows method, we direct our search more specifically for certain types of opening conditions that matches closes to the lowest Trough and to look for closing conditions that matches nearest to the highest peak.

When we say an EA is over optimised,  what we are generally saying is that the results bear little statistical significance and is an outcome  of random chance that doesn't produce consistency in performance and hence, unreliable.  So the problem doesn't lies in how many times you optimize but rather what's the rationale or method employed in the process of optimization, ie HOW you optimize.  Of cos the number of times you optimize does have an impact on the result outcome, but my discussion is focused on the process or method of optimization.