I agree Joe.  What I am saying though is that they do not "maintain continuity" 
of the equity line.  The fact that the equity amount repeats on the rollover 
day may not seem like a big deal but it actually compounds into a pretty 
serious discrepancy with a more frequent/shorter test periods.

--- In [email protected], "j0etr4der" <j0etr4...@...> wrote:
>
> From the Help section titled, " Walk-forward testing and optimization"
> 
> IN-SAMPLE and OUT-OF-SAMPLE combined equity
> 
> Combined in-sample and out-sample equities are available by ~~~ISEQUITY and 
> ~~~OSEQUITY composite tickers (consecutive periods of IS and OOS are 
> concatenated and scaled to maintain continuity of equity line - this approach 
> assumes that you generally speaking are compounding profits).
> 
> ...
> 
> I for one, would like a Setting on the Walk-Forward tab that disables this 
> behavior.
> 
> 
> --- In [email protected], "bh.hicks" <bh.hicks@> wrote:
> >
> > I've got some additional info on this.  When I compare a daily equity curve 
> > summary of the regular backtest against the WF OOS equity, it appears that 
> > the equity amount is repeating at the rollover period for the walk-forward 
> > version.  Regular backtest equity on left, walk-forward equity on right....
> > 
> > 12/29/1999  $233,209.95             12/29/1999      $233,209.95
> > 12/30/1999  $234,155.41             12/30/1999      $234,155.41
> > 12/31/1999  $233,625.53             12/31/1999      $233,625.53
> > 1/3/2000    $233,625.53             1/3/2000        $233,625.53
> > 1/4/2000    $229,524.14             1/4/2000        $229,524.14
> > 1/5/2000    $230,027.78             1/5/2000        $230,027.78
> > 
> > This error continues to compound at every walk-forward transition period 
> > and gets rather large by the end of the test.  I'm trying to resolve this 
> > with advanced WF period settings but has anyone else encountered this or 
> > has a work-around? Thanks.
> > 
> > --- In [email protected], "bh.hicks" <bh.hicks@> wrote:
> > >
> > > I've been trouble shooting this all day and am about to pull out my 
> > > remaining few hairs.  Any insight into this would be extremely 
> > > appreciated.
> > > 
> > > A walk-forward test on a system that has "dummy" variables should in 
> > > theory, match the results of a simple backtest over the same test period 
> > > as the variables aren't doing anything.  The same exact code is being run 
> > > across each period of time.  They aren't matching up.  
> > > 
> > > Fine you say, it is because of trades being exited at the end of each OOS 
> > > period vs trades being carried through those transition periods.  Nope, I 
> > > added a line to the code that exits all trades the last day of the year 
> > > and I am using 1/1/xx-12/31/xx OOS periods so all open trades are getting 
> > > dumped at the same time for regular backtest and WF.
> > > 
> > > Next step was to run regular backtests over my individual OOS WF periods. 
> > > Yes they match.  What I noticed here though is that since I am using 
> > > $100k as my initial equity, that means each WF step must also be using 
> > > $100k so the ~~~OSequity curve at the completion of a WF test is somehow 
> > > synthetically generated from a bunch of unique $100k segments rather than 
> > > carrying over the final equity from the previous OOS segment.  Is this 
> > > correct?  I suspect this is where the discrepancies are coming from but I 
> > > need to know for sure to have confidence in what I am seeing.  
> > > 
> > > Perhaps it has something to do with this compounding of $100k segments 
> > > rather than using the previous OOS segment's final equity as the starting 
> > > equity for the new segment?
> > >
> >
>


Reply via email to