Why backtests are useless, EAs are flawed and their parameters are bad [DISCUSS] | Currency Futures


futures.io - futures trading strategies, market news, trading charts and platforms


Currency Futures


This subforum allows you to find traders that trade Currency Futures. You may post about anything to do with trading currency futures in this section




 

Why backtests are useless, EAs are flawed and their parameters are bad [DISCUSS]

  #8 (permalink)

Regensburg, Germany
 
 
Posts: 9 since Oct 2013
Thanks: 0 given, 3 received


crazybears View Post
hi

only to add some very basic info about the topic .

I'd like to see more info about WFA , i made some TS and made WFA with AB and MC but wasn't able to made them work

Hey. I will make a new thread, before releasing my tool, that will explain how a WFA is done. Tough, its not the topic of this one

TS=Trading System; MC=Monte Carlo; ?
What is AB?

"Wasn't able to make them work"

Well, as long as you have a standardt EA, it will be just a few clicks to make a walk forward analysis with my tool.
So this sould not be a problem soon

To your Book suggestion:
I do not think there is any technique in any book that covers a good WFA, as a good WFA can only be done using program code, as there are up to 1000 optimisations and 100.000.000 backtests needed to get the whole picture. (I said UP TO, 100 optimisations and 100.000 backtests are more realistic).
Afterwards, this huge amount of data needs to be processed to find the ideal parameters...
So no human is able to do this in a reasonable time.

And I think a book that is only usable by coders would not be a bestseller xD

Btw: I am currently working on such an extended wfa algorithm, I will release it here in some months (free, yes. open source, i dont know) Stay tuned!


Last edited by Darwin; November 7th, 2013 at 11:33 AM.