Re: [Quantlib-dev] [QuantLib-svn] SF.net SVN: quantlib: [14166] trunk/QuantLib

[Ferdinando A. <na...@am...>]

> >          Volatility impliedVolatility(
> >                                Real price,
> >                                const Handle<YieldTermStructure>& discountCurve,
> > +                              Volatility guess,
> >                                Real accuracy = 1.0e-4,
> > -                              Size maxEvaluations = 100,
> > +                              Natural maxEvaluations = 100,
> >                                Volatility minVol = 1.0e-7,
> >                                Volatility maxVol = 4.0) const;
>
> Same concern here. The accuracy seems to me more likely to be specified
> than the guess. Moreover, the guess should have a default value, too.

I disagree here. In the previous implementation the guess was
hardcoded in the cpp file to be 10%. Providing a reasonable guess does
speed up computation significantly, and it seems to me uncontroversial
that a user might provide a reasonable guess, while usually he should
not specify accuracy: if he does, it's the accuracy default value
which is probably wrong... but that's for another possible future
change: moving from absolute accuracy to nominal-relative accuracy in
implied vol for swaption and cap/floor

As usual, despite being this uncontroversial for me I'm willing to
follow your suggestions if you're strong about them :-)

I'm agnostic about providing a default value for the guess, just
taught that with no default value there would be more chances of end
user noticing that he's now able to provide a guess

ciao -- Nando