OK.  I filter out securities that do not trade every day, so I do not have that 
problem.

BTW, I don't think you are rebalancing with the code below.  You are allowing 
the system to trade only on the last day of the month, but it is not 
rebalancing.

--- In [email protected], "Paolo" <pcavat...@...> wrote:
>
> I give you a quick example that can help.
> 
> If you want to rebalance at month-end you can code it as below
> 
> PositionScore = IIf(Month() != Ref(Month(), 1), myConditions, scoreNoRotate);
> 
> When using Rotational mode it will check the ranking - positionscore - on the 
> last day of every month. The point is that some of your securities may have 
> not been trading on that particular day and your ranking wouldn't be properly 
> calculated causing erroneous rotation.
> 
> Regular Backtest on the other side is working because doesn't require 
> comparing the ranking on the very same day accross all securities but you can 
> calculate the ranking at month end FOR EACH SECURITY no matter which was the 
> very last trading day for each of them.
> 
> For instance using:
> 
> CorrectPosition = ValueWhen(Month() != Ref(Month(), 1), myConditions);
> 
> and use it for proper buy/sell rules.
> 
> I hope it make it clearer.
> 
> Paolo
> 
> 
> --- In [email protected], "re_rowland" <rowland@> wrote:
> >
> > 
> > If you have data holes, then I don't see how not using rotational mode 
> > fixes your problem. Perhaps I don't understand your problem because I'm not 
> > sure what you mean by "month-end rebalancing signals are incorrectly 
> > interpolated."
> > 
> > --- In [email protected], "Paolo" <pcavatore@> wrote:
> > >
> > > Just wondering if anyone has ever noticed that backtestRotational doesn't 
> > > work properly when dealing with data holes.
> > > 
> > > These are quite frequent for instance at month-end across different 
> > > international asset classes.
> > > 
> > > Using the "Pad and align to reference symbol" doesn't always fix it 
> > > properly since month-end rebalancing signals are incorrectly interpolated.
> > > 
> > > This is just the last issue I've had with backtestRotational. My general 
> > > experience is that backtestRotational is theoretically very useful but I 
> > > often have to go back to code it using regular Backtest to make it work 
> > > the way I need.
> > > In this case month-end signals are correct no matter of data holes.
> > > 
> > > Just curious to know others' feedback.
> > > 
> > > Paolo
> > >
> >
>


Reply via email to