Hello,

Either sizing or adding offset is required if you want CONTINUOUS equity curve, 
otherwise at the end of one part of WF you would
have a jump back to initial equity. Choosing the model depends on using 
compounding or not.
Most users were interested in using compounding , hence this method is 
implemented. If you want offseting as an option use Feedback Center
to fill the suggestion.

Best regards,
Tomasz Janeczko
amibroker.com
  ----- Original Message ----- 
  From: Leading Edge Systems 
  To: [email protected] 
  Sent: Wednesday, July 22, 2009 9:41 PM
  Subject: [amibroker] Re: Walk Forward Result Not Accurate?




  I am new to AB but very familiar with walk forward analysis. AB should have 
an option to show the OSEQ curve without applying a sizing method so that the 
walk forward (which is usually done without sizing) net profit from each piece 
when added together will equal the OSEQ curve. It should also have the option 
of applying various sizing methods to the OSEQ set of trades. Do these options 
exist in AB?

  Thanks

  --- In [email protected], "jacklweinberg" <jacklweinb...@...> wrote:
  >
  > As I discussed privately with Thomas Ludwig, the key point is that the Walk 
Forward results that are shown in the spreadsheet at the end of the walk 
forward analysis, do NOT agree with the ~~~OSEQUITY ticker.
  > 
  > The reason is because the ~~~OSEQUITY ticker is "stitched" together from 
the out of sample equity tickers to be a continuous vector, as is explained in 
the AmiBroker documentation. That is compounding, and other factors are 
included in the ~~~OSEQUITY ticker.
  > 
  > HOWEVER, this implies that summary statistics based on the ~~~OSEQUITY 
ticker may or may not represent reality. 
  > For example, I have been using the summary stats (e.g. K-Ratio, Max 
Drawdown, etc.) that came from the ~~~OSEQUITY curve to compare the walk 
forward results of systems. 
  > 
  > I suppose that comparing systems this way is still reasonable. However the 
absolute values of the summary stats are somewhat in question.
  >



  

Reply via email to