[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [amibroker] Re: Optimization -- again



PureBytes Links

Trading Reference Links




<SPAN 
class=826570615-17102003>not to be flip, but what's the alternative? systems 
with no parameters? settings based on gut feelings, anecdotes or logic? whether 
we do it ourselves or our system does it mechanically by itself, isn't that how 
we decide what systems and settings are useful?
<SPAN 
class=826570615-17102003> 
<SPAN 
class=826570615-17102003>re a practical application, I have a framework I've 
been using to work on auto-optimization, an evolution of code I posted for this 
a few days ago. you can plug pretty much any trading logic into it, and the 
framework will automatically choose new settings every bar, for one or two 
parameters so far, based on what performed best over a lookback period you 
choose. I'm planning to post it, probably to the library so it's more readable 
than what happens to text in email, but it needs a little more documentation on 
how to set it up. bug reports and suggestions will be most 
welcome.
<SPAN 
class=826570615-17102003> 
<SPAN 
class=826570615-17102003>have I found that you can plug pretty much any 
common indicator into it and have it magically turn a nice profit? no way, in 
fact almost nothing I've tried this way is worth writing home about at all. 

<SPAN 
class=826570615-17102003> 
<SPAN 
class=826570615-17102003>that's why I keep banging on the theory and practice of 
optimization:
<SPAN 
class=826570615-17102003> 
<SPAN 
class=826570615-17102003>  - why doesn't this 
work?
<SPAN 
class=826570615-17102003> 
<SPAN 
class=826570615-17102003>  - if optimizing past returns doesn't 
produce good performance in the future, on what basis do we design and choose 
trading systems and parameters?
<SPAN 
class=826570615-17102003> 
<SPAN 
class=826570615-17102003>dave
<SPAN 
class=826570615-17102003> 
<BLOCKQUOTE 
>
  <FONT 
  size=2>From: Fred [mailto:fctonetti@xxxxxxxxx]<SPAN 
  class=826570615-17102003><FONT face="Courier New" 
  color=#0000ff> 
  Walk forward testing 
  and optimization is a great concept in theory and although I've seen lots 
  of ideas for how to set it up and be used over the years, unfortunately 
  I've yet to see anyone demonstrate that it actually works over a variety 
  of market conditions.  There are I'm sure lots of reasons for this 
  which I won't delve into here but the question remains, has anyone 
  actually seen this put into practice where the result has been a viable 
  system to use in mechanical trading ?  If so can you please point at 
  something that could be looked at objectively that doesn't reside in a 
  black box ?






Yahoo! Groups Sponsor












Send BUG REPORTS to bugs@xxxxxxxxxxxxx
Send SUGGESTIONS to suggest@xxxxxxxxxxxxx
-----------------------------------------
Post AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx 
(Web page: http://groups.yahoo.com/group/amiquote/messages/)
--------------------------------------------
Check group FAQ at: http://groups.yahoo.com/group/amibroker/files/groupfaq.html



Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.