1 Topic by hannahis (edited by hannahis 2018-09-21 09:13:54)

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

Topic: FSB Pro Optimization Sequence

Hi Popov,

The Strategy Rule section determines the Strategy's Entry accuracy and Exit Accuracy.  The Protection Section impact the Strategy's Exit Rule.

Which section has greater impact on the outcome of producing more Robust Strategy?

What make a strategy robust?  Which section has a greater impact on a strategy's robustness? The strategy's rules (opening/closing) or the money protection/Protection Section (SL, TP and BE)? (though both are important, it's the strategy's rules that try to find a repeatable patterns whereas, the money protection is finding a statistical/optimal exit with no bearing on repeatable patterns).


Under the Optimization Tool
I noticed that when I want to optimize the Strategy, the Optimizer seem to start optimizing the Protection Section (Stop Loss, Take Profit and Breakeven) 1st and then move on to optimize the Strategy's rules.  If this is observation is correct, could you kindly consider changing the optimization procedure.  Optimize the Strategy's Rules 1st and then optimize the Protection Section last.


In another words, can you start with optimize the Strategy's rule 1st, starting with the Biggest Time frame 1st, if LTF is being used and followed by the indicator that has the largest parameter value.  For eg if I have 2 H4 indicator rules, eg 1st rule - MA150 fast higher than slow MA200 for H4 and 2nd rule is MA100 fast higher than slow MA200 for H4, then optimize the bigger parameter 1st, i.e. MA150/200 then followed by the smaller parameter i.e. MA100/200 and if there are any lower time frame, use the same rule,

So we prioritize the optimization algorithm, Optimize 1) the Biggest Time frame 1st (eg. H4, H1, M30 in this sequence) , and if there are more than 1 rules that use the same time frame, then,

2) Optimize the indicator with the biggest Parameter value 1st and then move to the next 2nd highest parameter value

Once we have finished optimized the indicators rules, then move on to optimize the Protection Section (Stop Loss, Take Profit and Break Even).

The reason why we need to optimize the Indicator Rules 1st and not the Protection Section because it's the Strategy's rule that need to be refined 1st, i.e. find the best parameter setting that produce more robust reaction to the market fluctuation.

It's the rules that make the Strategy more robust, the SL, TP, BE do not constitute to better rules, it only improve the strategy's exit accuracy and not the strategy's entry rules.  A good EA is 1st of all, good entry then followed by good exit.

A good entry is more important than a good exit (I'm not saying it's not important but lesser in comparison though both are still very important).


For those who just want to refine your exit rules only in the Protection Section, we can then un-check/tick all the indicators sections. 

For those who only want to refine the entry/exit rules only, then likewise, un-check/tick the Protection Section. 

The benefit of choosing which section you want to optimize 1st would have you focus which section (indicator/protection) you are aiming to improve 1st.  However, this method isn't complete because we can't optimize both together in the manner of focusing improving the indicators first and then improve the protection section later.  However, this method means we have to break up the optimizing time/procedure into 2 separate efforts instead of 1 seamless procedure.

Ideally, if we can have the option to choose which to optimize 1st, that would be perfect.  And we get to optimize both sections in 1 procedure (but with the choice to choose which section to optimize 1st)

2 Reply by GD (edited by GD 2018-09-21 13:17:52)

  • GD
  • GD
  • Licensed Member
  • Offline
  • Registered: 2015-05-22
  • Posts: 444

Re: FSB Pro Optimization Sequence

We can do that with FSB Pro.

  • optimization with no protection parameters permitted

  • optimization of protection parameters by keeping constant all the other parameters


1. But how we can do it with EA Studio?
2. What you say has meaning as some functions can have a lot of minima areas in the poly parametric space. So the procedure to find one of them depends on way to find them. BUT which minimum or maximum of statistical function gives more robust EA is still not clear. If so, then...

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

Re: FSB Pro Optimization Sequence

GD wrote:

We can do that with FSB Pro.

  • optimization with no protection parameters permitted

  • optimization of protection parameters by keeping constant all the other parameters

That's what I meant when I wrote "For those who just want to refine your exit rules only in the Protection Section, we can then un-check/tick all the indicators sections.

For those who only want to refine the entry/exit rules only, then likewise, un-check/tick the Protection Section...However, this method means we have to break up the optimizing time/procedure into 2 separate efforts instead of 1 seamless procedure."


GD wrote:

What you say has meaning as some functions can have a lot of minima areas in the poly parametric space. So the procedure to find one of them depends on way to find them. BUT which minimum or maximum of statistical function gives more robust EA is still not clear. If so, then...

I'm not sure whether my explanation below is what you are asking about...

The reason I suggest we optimize the Bigger/highest parameter 1st is the same concept I suggest that when we build our EA,input the Higher time frame first and then follow by the lower time frame.

In another words, We use eg H4/D1 indicator to determine the trend and the lower time frame to trigger the entry.  So the bigger parameter is liken the Higher Time frame to determine the trend and the lower parameter to improve the entry accuracy.

4 Reply by GD (edited by GD 2018-09-22 05:58:49)

  • GD
  • GD
  • Licensed Member
  • Offline
  • Registered: 2015-05-22
  • Posts: 444

Re: FSB Pro Optimization Sequence

Like preset indicators in EA Studio, with larger values for parameters.
In this case we can use EA studio in higher TF to find parameters of indicators and finally use them as preset indicators at lower ones.

5 Reply by hannahis (edited by hannahis 2018-09-22 08:39:42)

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

Re: FSB Pro Optimization Sequence

Yes, you are right.  By 1st optimizing the Biggest parameter 1st or the Higher Time Frame 1st, we are looking for a bigger or more stable trend movement and once this trend movement is found, we then move on to re-fine/optimize the lower time frame/parameter.  So thus, using the biggest parameter/time frame as a "preset" indicator to guide us in the rest of the optimization process.

I'm not sure how the current Optimization algorithm works/the rationale or thinking behind it.  For example, when it is optimizing the 1st rule, does it also take into account of the rest of the rules (eg 2nd rule, 3rd rule, 4th rule etc remain in the equation when optimizing?).  If yes, then it's like trying to optimize the Biggest rule 1st to "fit" well with the 2nd, 3rd and 4th rule....which is not what I'm proposing.

I'm hoping somehow the algorithm would optimize the Biggest Time Frame 1st (without taking into account of the rest of the rules, as if that's the only rule and then move on to add the 2nd rule (the 2nd highest time frame/parameter) and optimize the 2nd rule to fit the 1st rule (instead of optimizing the 1st rule to fit the rest of the rules).

So it's like optimizing rule by rule, layer by layer.

In equation form, it's like Optimize Rule 1 (which is the Biggest Time Frame/parameter).  Equation = Optimize 1st rule only and once 1st rule most optimal parameters are found, add in the 2nd rule and optimize the 2nd rule to fit best with the 1st rule.

Not Optimize 1st rule while other rules are still in the equation eg Equation = 1st rule (optimize), while 2nd rule, 3rd rule and 4th rule are also taken into consideration as part of the equation (so that all the 4 rules together = profitable strategy.  The problem is, what if the 2nd rule and 3rd rules are lousy rules and we are trying to optimize the 1st rule to fit into those remaining lousy rules and we ended up with a lousy 1st rule.