Live trades for NN EA idea - page 14

 
leeb:
Another successful trade, equity now at $15,826

hi, Leeb

i appreciate your equity line using ea, for now it seems good ...

i am continuing to follow your result

giapel

 

Well the system made a few big losses in a row, it had too many lots on the table at once but I think the logic behind it is sound. Need to tweak money management. I might share some of the thinking behind the EA if anyone would like to discuss though this won't include code but I think could lead to some interesting new systems. Will be closing the platform this EA was running on now to tweak the MM.

 
 

The thinking behind the system

Hi Mic,

Thanks for your input. It makes a lot of sense what you say and the questions you ask.

So to reveal the way of thinking behind the EA... maybe some people can experimemt with this new (don't think it has been discussed here before) and we could start a new thread. The way of thinking is that the EA is trained on several different months of data, eg: subsystem 1 is optimised to 1 months data, subsystem 2 to 3 months of data, subsystem 3 to 6 months of data... what do you all think to this approach ? It should be much more dynamice than just optimising to the last month of data ? Any feedback welcome, maybe we can test optimising some EAs with this approach such as Phoenix and start a new thread...

 

For a MT base NN system, in theory the longer period of data to optimize the NN is aiming for the long term and the shorter period of data is aming for short term both have their own beauties. The shortest optimization period of my test NN system is 15 days, it gives good one to two days forecast, but it needs optimization more often then the longer period.

May be THIS could save some time for regular optimization, don’t know if any of you have try this before?

Mic.

 

Hi mic, sounds like we are thinking along very similar lines, using different periods of time to optimise I tried the link you provided re: auto-optimising in MT4 but the EA never placed any trades..

 
 

Hi,

i have been following this thread for awhile.. i am also studying to make my own EA.. this is what i have in mind..

maybe we can find someone to create a auto-optimize mqh.. with more settings than what original auto-optimize offer like more variables, and also limitation like max drawdown

i think any system can have a better outcome if it changes settings with the market conditions..

Regards,

Antone

 
leeb:
Hi Mic,

Thanks for your input. It makes a lot of sense what you say and the questions you ask.

So to reveal the way of thinking behind the EA... maybe some people can experimemt with this new (don't think it has been discussed here before) and we could start a new thread. The way of thinking is that the EA is trained on several different months of data, eg: subsystem 1 is optimised to 1 months data, subsystem 2 to 3 months of data, subsystem 3 to 6 months of data... what do you all think to this approach ? It should be much more dynamice than just optimising to the last month of data ? Any feedback welcome, maybe we can test optimising some EAs with this approach such as Phoenix and start a new thread...

I am doing something like your system

I have a few question:

Is your NN system train data base on some exits system (somethine like macd EA system shipped with mt4)

how can I make NN system better ?

Is it possible make a high probility system by NN?