1 (edited by Naya 2022-08-22 15:31:54)

Topic: Expand FSB Pro Collection beyond 100 strategies

Hi Popov,
In my testing of EA studio I noticed it's possible to download the settings file in .JSON format and edit the collection capacity to any number you wish and upload the settings file again and EA studio will work with it. With that I could create a collection of 20000 strategies and more.
Is it possible to do such in FSB pro?
My experience with FSB pro has been to save multiple collections with a 100 strategies as the generator runs. Having a larger collection capacity will make the work easier.

(* Sometimes the FSBgenerator 'accidentally' sends 101strats to a collections and the software gives an out of bounds error. It seems it's a bug. When this occurs, i  would delete the extra strat before saving the collection)

Thanks

Re: Expand FSB Pro Collection beyond 100 strategies

Hi Naya, I'm trying to test some collections of 100 strategies of the EA studio in the FSB Pro, but i can't compile all these strategies to put in the FSB. How did you export these collections to FSB to test then ?

Re: Expand FSB Pro Collection beyond 100 strategies

> Sometimes the FSBgenerator 'accidentally' sends 101 strats to a collection, and the software gives an out-of-bounds error.

Thank you for mentioning this issue. We will inspect it.

> Hi Naya, I'm trying to test some collections of 100 strategies of the EA studio in the FSB Pro,

We are working on providing full support of EA Studio collections in FSB Pro.

4 (edited by sleytus 2022-08-23 17:28:06)

Re: Expand FSB Pro Collection beyond 100 strategies

Naya wrote:

Hi Popov,
In my testing of EA studio I noticed it's possible to download the settings file in .JSON format and edit the collection capacity to any number you wish and upload the settings file again and EA studio will work with it. With that I could create a collection of 20000 strategies and more.
Is it possible to do such in FSB pro?
My experience with FSB pro has been to save multiple collections with a 100 strategies as the generator runs. Having a larger collection capacity will make the work easier.

(* Sometimes the FSBgenerator 'accidentally' sends 101strats to a collections and the software gives an out of bounds error. It seems it's a bug. When this occurs, i  would delete the extra strat before saving the collection)

Thanks

I'm not clear about what you are attempting to achieve -- do you really plan on trading 20000 individual strategies or a portfolio of 20000 strategies?  Or are you just playing?  Have you ever traded a live account?

How exactly does saving multiple collections with 100 strategies "make the work easier"?  What "work" are you referring to?

Here you have an amazing tool at your finger tips -- and instead of focusing how to best use it in ways it was designed and intended, you are attempting to twist it to adapt to a misconception.  Users of a software tool should be prepared to adapt to the tool, not the other way around.  Before attempting to change the tool, how about first learning to use it to trade profitably?  I mean, that's the ultimate goal, right?  The goal is to trade, right?

Re: Expand FSB Pro Collection beyond 100 strategies

Juantele wrote:

Hi Naya, I'm trying to test some collections of 100 strategies of the EA studio in the FSB Pro, but i can't compile all these strategies to put in the FSB. How did you export these collections to FSB to test then ?

EA Studio and FSB Pro are separate tools.  As a new user they are both unfamiliar to you.  And though you are not familiar with either one, the first thing you want to do is try to mix them?  Why not learn to use one (or both) to create quality strategies that you can trade?  I mean, the goal is to profitably trade, right?

Re: Expand FSB Pro Collection beyond 100 strategies

That work is data mining and huge amounts of data (strategies) are needed for it. If we take Studio as an example, then it is most useful for data mining and not for model based development.

sleytus wrote:
Naya wrote:

Hi Popov,
In my testing of EA studio I noticed it's possible to download the settings file in .JSON format and edit the collection capacity to any number you wish and upload the settings file again and EA studio will work with it. With that I could create a collection of 20000 strategies and more.
Is it possible to do such in FSB pro?
My experience with FSB pro has been to save multiple collections with a 100 strategies as the generator runs. Having a larger collection capacity will make the work easier.

(* Sometimes the FSBgenerator 'accidentally' sends 101strats to a collections and the software gives an out of bounds error. It seems it's a bug. When this occurs, i  would delete the extra strat before saving the collection)

Thanks

I'm not clear about what you are attempting to achieve -- do you really plan on trading 20000 individual strategies or a portfolio of 20000 strategies?  Or are you just playing?  Have you ever traded a live account?

How exactly does saving multiple collections with 100 strategies "make the work easier"?  What "work" are you referring to?

Here you have an amazing tool at your finger tips -- and instead of focusing how to best use it in ways it was designed and intended, you are attempting to twist it to adapt to a misconception.  Users of a software tool should be prepared to adapt to the tool, not the other way around.  Before attempting to change the tool, how about first learning to use it to trade profitably?  I mean, that's the ultimate goal, right?  The goal is to trade, right?

7 (edited by sleytus 2022-08-23 17:59:04)

Re: Expand FSB Pro Collection beyond 100 strategies

What do you mean by "data mining"?  There is only a single data set -- i.e. the same data horizon that is used by all the algebraic formulas embedded in strategies.  The term "data mining" to me implies there is more data to be had.

Also, how many strategies can one possibly trade (or manage)?  When it comes to strategies -- is the goal quality or quantity?

Suppose you have 20000 strategies -- are you going to trade all of them?  Are you going to prune them to the best 100?  What criteria will you use for pruning?  Visual inspection? 



footon wrote:

That work is data mining and huge amounts of data (strategies) are needed for it. If we take Studio as an example, then it is most useful for data mining and not for model based development.

sleytus wrote:
Naya wrote:

Hi Popov,
In my testing of EA studio I noticed it's possible to download the settings file in .JSON format and edit the collection capacity to any number you wish and upload the settings file again and EA studio will work with it. With that I could create a collection of 20000 strategies and more.
Is it possible to do such in FSB pro?
My experience with FSB pro has been to save multiple collections with a 100 strategies as the generator runs. Having a larger collection capacity will make the work easier.

(* Sometimes the FSBgenerator 'accidentally' sends 101strats to a collections and the software gives an out of bounds error. It seems it's a bug. When this occurs, i  would delete the extra strat before saving the collection)

Thanks

I'm not clear about what you are attempting to achieve -- do you really plan on trading 20000 individual strategies or a portfolio of 20000 strategies?  Or are you just playing?  Have you ever traded a live account?

How exactly does saving multiple collections with 100 strategies "make the work easier"?  What "work" are you referring to?

Here you have an amazing tool at your finger tips -- and instead of focusing how to best use it in ways it was designed and intended, you are attempting to twist it to adapt to a misconception.  Users of a software tool should be prepared to adapt to the tool, not the other way around.  Before attempting to change the tool, how about first learning to use it to trade profitably?  I mean, that's the ultimate goal, right?  The goal is to trade, right?

Re: Expand FSB Pro Collection beyond 100 strategies

As an aside -- it is my opinion that you can take strategy-generation only so far.  The vast majority of indicators (and EAs) are MA-based (i.e. based on Moving Average algorithms).  And the most important setting for an MA-based algorithm is 'Period'.  By fiddling with input settings one can get almost any MA-based algorithm to generate signals similar to another MA-based algorithm.  Though it appears there are many ways to combine settings, in the end there aren't that many combinations that would yield consistent results.

On the other hand, if you really want to take your trading to the next level then the focus should be on "Money Management" (MM).  MM is where "the rubber meets the road".  MM is not well-defined and means different things to different people -- so, it's rarely discussed in meaningful ways.  Again -- this is my opinion.

In the case of EA Studio and FSB Pro the only elements that I would consider that fall under the domain of MM are StopLoss / TakeProfit and (maybe) Trailing Stop.  Another area of MM that, as far as I can tell, is not yet explored by these tools, is what action to take when trades are lost or DD reaches a particular threshold.  There is also "dynamic pruning" -- where strategies are toggled ON/OFF based on how they are currently performing.

Re: Expand FSB Pro Collection beyond 100 strategies

You mine strategies because not all strats are equal. The more you have them, the bigger the chance to find profitable ones.

There are many parameters for MA and don't forget generator is combining indis, so the number of combinations is not low to say the least.

20k strats is just the beginning, this is the pool of strats, which needs to get filtered, manipulated or whatever one's methodology demands.

sleytus wrote:

What do you mean by "data mining"?  There is only a single data set -- i.e. the same data horizon that is used by all the algebraic formulas embedded in strategies.  The term "data mining" to me implies there is more data to be had.

Also, how many strategies can one possibly trade (or manage)?  When it comes to strategies -- is the goal quality or quantity?

Suppose you have 20000 strategies -- are you going to trade all of them?  Are you going to prune them to the best 100?  What criteria will you use for pruning?  Visual inspection?

Re: Expand FSB Pro Collection beyond 100 strategies

sleytus wrote:
Juantele wrote:

Hi Naya, I'm trying to test some collections of 100 strategies of the EA studio in the FSB Pro, but i can't compile all these strategies to put in the FSB. How did you export these collections to FSB to test then ?

EA Studio and FSB Pro are separate tools.  As a new user they are both unfamiliar to you.  And though you are not familiar with either one, the first thing you want to do is try to mix them?  Why not learn to use one (or both) to create quality strategies that you can trade?  I mean, the goal is to profitably trade, right?

hi, i wanted to pick the collections generated in ea studio and put than in other data to test in fsb to make the work more efficient

11 (edited by sleytus 2022-08-24 02:46:37)

Re: Expand FSB Pro Collection beyond 100 strategies

Juantele wrote:
sleytus wrote:
Juantele wrote:

Hi Naya, I'm trying to test some collections of 100 strategies of the EA studio in the FSB Pro, but i can't compile all these strategies to put in the FSB. How did you export these collections to FSB to test then ?

EA Studio and FSB Pro are separate tools.  As a new user they are both unfamiliar to you.  And though you are not familiar with either one, the first thing you want to do is try to mix them?  Why not learn to use one (or both) to create quality strategies that you can trade?  I mean, the goal is to profitably trade, right?

hi, i wanted to pick the collections generated in ea studio and put than in other data to test in fsb to make the work more efficient


Why would you need to test the collections generated by EA Studio in FSB Pro?  Even if you could import EA Studio collections into FSB Pro, do you have any evidence or experience that FSB Pro could improve them?  If you think FSB Pro creates better strategies then why bother with EA Studio -- simply use FSB Pro.  I'm not recommending that you abandon EA Studio -- just that I'm suggesting before you attempt to use the tools in ways they weren't intended, how about first learning how to use the tools in the ways they were designed to be used.

Popov has gone to great lengths to provide excellent documentation and videos -- maybe check some of those out?

I'm not trying to give you a hard time -- rather, my point is you will be better served if you learn to use the tools in the way they were designed rather than attempting to twist them in ways that you want them to work.

Re: Expand FSB Pro Collection beyond 100 strategies

Juantele wrote:
sleytus wrote:
Juantele wrote:

Hi Naya, I'm trying to test some collections of 100 strategies of the EA studio in the FSB Pro, but i can't compile all these strategies to put in the FSB. How did you export these collections to FSB to test then ?

EA Studio and FSB Pro are separate tools.  As a new user they are both unfamiliar to you.  And though you are not familiar with either one, the first thing you want to do is try to mix them?  Why not learn to use one (or both) to create quality strategies that you can trade?  I mean, the goal is to profitably trade, right?

hi, i wanted to pick the collections generated in ea studio and put than in other data to test in fsb to make the work more efficient

Add new data to Studio, you can have multiple sources for the same timeframe, by doing it this way you'll win in efficiency.

13 (edited by sleytus 2022-08-24 13:52:41)

Re: Expand FSB Pro Collection beyond 100 strategies

footon wrote:

You mine strategies because not all strats are equal. The more you have them, the bigger the chance to find profitable ones.

There are many parameters for MA and don't forget generator is combining indis, so the number of combinations is not low to say the least.

20k strats is just the beginning, this is the pool of strats, which needs to get filtered, manipulated or whatever one's methodology demands.

sleytus wrote:

What do you mean by "data mining"?  There is only a single data set -- i.e. the same data horizon that is used by all the algebraic formulas embedded in strategies.  The term "data mining" to me implies there is more data to be had.

Also, how many strategies can one possibly trade (or manage)?  When it comes to strategies -- is the goal quality or quantity?

Suppose you have 20000 strategies -- are you going to trade all of them?  Are you going to prune them to the best 100?  What criteria will you use for pruning?  Visual inspection?


Hi footon -- thanks for your response.  If you don't mind -- could we pursue this a bit further.  In my first go around with forexsb I learned a **lot** from you...

I understand how the generator combines indis, that there are zillions of possible combinations, and that they are not all created equal.  My comment about the commonality of MA-based indis is approximately true, though not 100% true.  Yes -- MA-based indis may include other settings besides 'Period', but by fiddling with the settings of one MA-based indi you can often get it to generate a pattern similar to the pattern of a different MA-based indi.  The point I was trying to make -- but seemed to fail -- is that it is less important to focus on the number of strategies than on the quality of strategies.   And, yes, being able to filter 20000 strats is better than only having 100 strats to filter -- however, running the Reactor longer or running the Validator multiple times (applying different filters) will get you to the same place.

I mean, in the end you are not going to trade 20000 strats -- rather, you want to whittle that down to a manageable number to trade -- e.g. 50.   You don't need 20000 strats to end up with 50 "quality" ones.  You can achieve a similar result by running the Reactor longer and / or running the Validator multiple times applying different filters to the same collection until you reach the goal of 50 good strategies -- and this doesn't require editing *.json files.

And, of course, in the end we should all know that though pretty statistics make us feel warm and fuzzy inside they really don't make a whole lot of difference.  The market will do what the market will do.  Market conditions constantly change and the data patterns tomorrow may be different than anything your algebra (i.e. strats) have ever seen.  Plus, live trading uses streaming tick data whereas backtesting (and indis) uses static OHLC data -- a big difference -- so, there will always be a disconnect between backtesting and demo / live trading.  That disconnect isn't solved by more strategies.  Furthermore, when trading a live account you have probably seen that often times strats with poorer stats perform better than strats with prettier stats.  So, though statistics are important, obsessing about them is a misuse of your brain power -- in my opinion...

Re: Expand FSB Pro Collection beyond 100 strategies

Of course, I know what you mean -- though I don't think this is necessarily true.  And that is because there is a point at which repeated refining of statistics has minimal or no affect on actual trading performance.  For example, suppose I create a portfolio of 100 strategies with amazing, amazing statistics.  And suppose I rank them from 1 to 100 based on one of their statistical metrics (I don't care -- choose a metric).  Would you claim that when you actually trade those strategies in a live account that they will perform according to the way you ranked them?  I highly doubt that -- in fact, I know that wouldn't be the case.  And that is because regardless how much effort we put into improving the backtesting statistics -- there will always be a disconnect between backtesting with static OHLC data compared to demo / live trading with streaming tick data. The only way to ensure that all the time we spend on backtesting and generating pretty statistics isn't for naught is if we had a time machine or crystal ball -- in my opinion.

footon wrote:

You mine strategies because not all strats are equal. The more you have them, the bigger the chance to find profitable ones.

Re: Expand FSB Pro Collection beyond 100 strategies

there will always be a disconnect between backtesting with static OHLC data compared to demo / live trading with streaming tick data

No, no, no, no, no, no smile Tick data is not what it's trading on, OHLC is still used in live trading! Live trading corresponds to backtesting behaviour, that is a fact, and it cannot be different, otherwise backtesting would lose its whole point. If we consider strategies with bar open as the opening point, then what's it got to do with ticks? Opening condition indis are still using OHLC for value calculations. If there is other opening point, like a MA, then on one hand position gets opened when price in the form of tick touches/crosses the line, on the other hand in backtesting it is still obvious at what price position gets opened, we just don't know the exact time but that it is not necessary, we know the bar time and price extremes during that bar to know possible outcomes.

Alright, I have to make a break because a thunder storm is getting close...

Re: Expand FSB Pro Collection beyond 100 strategies

You don't need 20000 strats to end up with 50 "quality" ones

This is methodology dependant. What if criteria or whatever is used for validation produces 1 quality strat out of a pool of 400? To get 50 Q-strats, one would need 20k strats.

of one MA-based indi you can often get it to generate a pattern similar to the pattern of a different MA-based indi

Indis that form a strategy need to sing a different tune, meaning their parametes are distinct to one another. That is what makes up a huge number of available combinations.

however, running the Reactor longer or running the Validator multiple times (applying different filters) will get you to the same place

Yes, one way or other the investment in time cannot be shortened.

Furthermore, when trading a live account you have probably seen that often times strats with poorer stats perform better than strats with prettier stats.  So, though statistics are important, obsessing about them is a misuse of your brain power

I agree. Years ago when Studio came out, I conducted a very labour-intensive research about the distribution of profitable strategies in collections. Fascinating results, but very dull manual work. I came to a conclusion that Q-strats are from quarter to middle in 100 strat collections, strategies ranked from ~25 - ~40 showed profitability, top of the pops broke barely into break-even territory. Would be interesting to see if this behaviour has carried over to today's strats and markets.

In my first go around with forexsb I learned a **lot** from you...

Thank you! I'll take it as a compliment but question everything!

Re: Expand FSB Pro Collection beyond 100 strategies

Hey footon -- no disagreement.  Thanks for your response...

Your comment about the distribution of profitable strategies in collections is very interesting.  I haven't done a methodical investigation into this the way you have, but I do have many trades under my belt and I would say they are consistent with your observations.  That is one reason why I'm inclined to suggest to newcomers they don't obsess over creating prettier and prettier statistics because there comes a point where prettier does not translate to better performance.  In fact, there may be a point where prettier suggests over curve-fitted.  This is where money management (MM) comes in.  MM is poorly defined and means something different to different people.  For me -- MM is all the logic (and code) that falls outside indicator algebra.  MM seems to be the one thing missing from Forexsb software -- perhaps a topic for another thread...


footon wrote:

You don't need 20000 strats to end up with 50 "quality" ones

This is methodology dependant. What if criteria or whatever is used for validation produces 1 quality strat out of a pool of 400? To get 50 Q-strats, one would need 20k strats.

of one MA-based indi you can often get it to generate a pattern similar to the pattern of a different MA-based indi

Indis that form a strategy need to sing a different tune, meaning their parametes are distinct to one another. That is what makes up a huge number of available combinations.

however, running the Reactor longer or running the Validator multiple times (applying different filters) will get you to the same place

Yes, one way or other the investment in time cannot be shortened.

Furthermore, when trading a live account you have probably seen that often times strats with poorer stats perform better than strats with prettier stats.  So, though statistics are important, obsessing about them is a misuse of your brain power

I agree. Years ago when Studio came out, I conducted a very labour-intensive research about the distribution of profitable strategies in collections. Fascinating results, but very dull manual work. I came to a conclusion that Q-strats are from quarter to middle in 100 strat collections, strategies ranked from ~25 - ~40 showed profitability, top of the pops broke barely into break-even territory. Would be interesting to see if this behaviour has carried over to today's strats and markets.

In my first go around with forexsb I learned a **lot** from you...

Thank you! I'll take it as a compliment but question everything!