Optimizing EAs in Real Time

 

Dear Programmers, Can anyone tell me how to optimize EAs in real time?

As no one has any clue why is this needed, I will post something very interesting.

That all programmers will dare to change their perception of trading and move to a new level of trading. 

 

 

How can you take the results from the optimizing process input it into a learning algorithm and applying it to your input variables for the expert?

 

 
PCWalker:

How can you take the results from the optimizing process input it into a learning algorithm and applying it to your input variables for the expert?

 

You don't.

You make a copy of the initial external variables, and then modify those copies realtime.

 
Marco vd Heijden:

You don't.

You make a copy of the initial external variables, and then modify those copies realtime.

how you get the initial variables in real time? and if you can't get the optimization process variables in real time. the problem is not solved.

Which Initial external variables? and what do you mean by modifying them in real time? How can you modify them in real time, when the optimization process takes 2-20 days?

 

The initial variables are the last best values known, or called all time variables.

These variables are calculated over the largest batch of available historic data.

These values define the bandwith between what alway's happens, what never happened and what is expected (what usually) happens.

The calculations are recalculated on a ongoing basis, a set interval or realtime.

Upon loading the system, the last known values are the last- most up to date outcomes.

Because input parameters are fixed (static) and can not be changed while the system is running, they are copied to a live set of variables and these variables are overwritten with new data whenever a change or difference arises between the known and calculated outcomes.

Modifications in real time will happen for example when there will be a new all time high or low, a new number in for example bar volume, or a new bar that exceeds all other known bars in amplitude, all time bars,and or related only to its time stamp.

It is not a given that the optimization process takes 2-20 days, in fact this number can vary between any time span from all available historic data to mere minutes, it all depends upon the system designer.

A mix of time blocks is also possible, first the all time variables are calculated over all available history, then over for example only the last year, the last week, yesterday and etc.

So to end up with even more bandwitdth settings, long term, short term and data for noise suppressors.

Its a wide subject, to create a self adjusting system, i can only set out the base lines here.

 
Marco vd Heijden:

The initial variables are the last best values known, or called all time variables.

These variables are calculated over the largest batch of available historic data.

These values define the bandwith between what alway's happens, what never happened and what is expected (what usually) happens.

The calculations are recalculated on a ongoing basis, a set interval or realtime.

Upon loading the system, the last known values are the last- most up to date outcomes.

Because input parameters are fixed (static) and can not be changed while the system is running, they are copied to a live set of variables and these variables are overwritten with new data whenever a change or difference arises between the known and calculated outcomes.

Modifications in real time will happen for example when there will be a new all time high or low, a new number in for example bar volume, or a new bar that exceeds all other known bars in amplitude, all time bars,and or related only to its time stamp.

It is not a given that the optimization process takes 2-20 days, in fact this number can vary between any time span from all available historic data to mere minutes, it all depends upon the system designer.

A mix of time blocks is also possible, first the all time variables are calculated over all available history, then over for example only the last year, the last week, yesterday and etc.

So to end up with even more bandwitdth settings, long term, short term and data for noise suppressors.

Its a wide subject, to create a self adjusting system, i can only set out the base lines here.

Can you show me in any text book of financial engineering or Financial Statistics, where you came up with all these ideas? By the way, if you can't show me any documentation of your initial variables, I am happy that you are a very creative person, and you like to create new theories. :-) hehehehe
 

Well if you think financial markets is about textbooks or that everything is written down somewhere or that every idea has to come out of a book, i do not know what to tell you.

I have read a lot of books but i can not point you to the exact phrases that describe the logical elements of (regular) market bandwidth.

It is only logical to play between these boundaries because if you use unrealistic values, or you do not know what the boundaries are then it will quickly resemble a guessing game.

I always tend to think the market has a specific frequency, and you have to tune your system to that frequency, just like you tune your radio to a specific station.

If your tunnning is off then the music will be of a bad quality.

I think the question to point to a specific reference is a wrong one.

These things are not templates these are things you find along the way when you decide to try things and you discard what doesn't work and you combine what works to end up with your best tunning mechanism.

This process takes years to refine and it will take maky tests and many idea's of various sources.

My collection of books is over a 1000+ and i do not have the correct memory, nor do i feel like looking up all of the elements to provide you with an exact blueprint of how i got to these formulas.

But i can tell you that time is the most important element, and time combined with calculations data about usual happenings, can be used to detect unusual happenings, and these can be traded with a certianity that i found hard to beat with any other system that ive tried so far.

If you would be asking for some of these calculations, in stead of text book references, that would be a different and most likely more productive story.

Because i can certainly show you how it's done, not where to find it.

 
Marco vd Heijden:

Well if you think financial markets is about textbooks or that everything is written down somewhere or that every idea has to come out of a book, i do not know what to tell you.

I have read a lot of books but i can not point you to the exact phrases that describe the logical elements of (regular) market bandwidth.

It is only logical to play between these boundaries because if you use unrealistic values, or you do not know what the boundaries are then it will quickly resemble a guessing game.

I always tend to think the market has a specific frequency, and you have to tune your system to that frequency, just like you tune your radio to a specific station.

If your tunnning is off then the music will be of a bad quality.

I think the question to point to a specific reference is a wrong one.

These things are not templates these are things you find along the way when you decide to try things and you discard what doesn't work and you combine what works to end up with your best tunning mechanism.

This process takes years to refine and it will take maky tests and many idea's of various sources.

My collection of books is over a 1000+ and i do not have the correct memory, nor do i feel like looking up all of the elements to provide you with an exact blueprint of how i got to these formulas.

But i can tell you that time is the most important element, and time combined with calculations data about usual happenings, can be used to detect unusual happenings, and these can be traded with a certianity that i found hard to beat with any other system that ive tried so far.

If you would be asking for some of these calculations, in stead of text book references, that would be a different and most likely more productive story.

Because i can certainly show you how it's done, not where to find it.

Thank you, But I don't think you really understand optimization process, or for that matter Applied Statistics to Optimization Data. If you just create ideas without common grounds or some basic common ideas, then I am quite sure, you are mostly successful at your belief, not at real Mathematics, Statistics, or Finance. I am happy for you.
 

Okay thats fine.

It saves me a lot of time.

Good luck.

 
Marco vd Heijden:

Okay thats fine.

It saves me a lot of time.

Good luck.

I am sorry, you are very clever, and you right, your beliefs and wisdom applies to your reality not to mine.

And I am quite sure even if I am 13 years old, you are making much more money then I am making, therefor MetaQuotes should pay you more for your precious time.

And Thank you for your time, and do try to save more of it, so that you can make more money in your life. 

 

I am only in shortage of time.

It is our most precious commodity, it can never be bought, or bought back.

Not sure what you mean by all the making money things i was under the impression you was going about trading systems and in particular self learning systems.

Reason: