Re: [Quantlib-dev] Adding new functions to blackcalculator

[Dima <dim...@go...>]

Ok, thanks a lot for your replies! After thinking about it I'd suggest
to code the delta calculations in a new class called BlackDeltaCalculator
with a similar constructor as the BlackCalculator class. This might look
like
overkill (an own class for Delta calculation) on first sight, but my reasons
are
as follows:

- we'll have to deal with 4 deltas, which could be conveniently
coded in an enumeration which could be passed to the
constructor.
- we'll need 4 different functions to return a strike for a given
delta. 2 of them will include a zero root search algorithm (premium
adjusted ones). This could be all handled by one function strike().
- We will need at least two functions, which return the strike for
an ATM convention (e.g. atm forward or atm delta neutral straddle).
Delta neutral straddle strike depends again on the 4 delta conventions.

So, to summarize: if we would try to build that in the current
BlackCalculator
class, this would mess things up a lot, since many delta specific functions
would be needed, which are mostly used in FX.

I'm very open to other suggestions. What do you think? Regards, Dima














2008/11/13 Ferdinando Ametrano <na...@am...>

> On Thu, Nov 13, 2008 at 12:42 PM, Dima <dim...@go...>
> wrote:
> > Ok, good. [...] So, how to proceed?
> It would be better if you work on the current trunk version, so
> familiarize yourself with svn and check out the trunk snapshot. See
> http://quantlib.org/svn.shtml
> Also please read the "Developer introduction" at
> http://quantlib.org/newdeveloper.shtml
>
> > Will I work with someone of the core
> > developers of the class closely?
> I am the class main developer, but a) I'm not that proud of its design
> b) it's not that complex c) Luigi is THE man
> Just post on the dev list and somebody will step up and help you.
>
> > Since I wasn't sure
> > what to do with the blackcalculator class.
> Please take a look at the blackformula file, where you can find
> (hopefully) efficient formula for price and implied vol. You might
> probably consider to implement strikeForDelta as function instead of
> BlackCalculator class method.
> BlackCalculator is just an aggregation of less used formulae with some
> common calculation factorized at construction time.
> See also BlackScholesCalculator. ImpliedVol and StrikeForDelta might
> be wrapped as BlackCalculator methods if needed
>
> > It seems,
> > like Asset or Nothing and Cash or Nothing options
> > are dealed with in the same class. But that is something
> > that would be difficult to incorporate for functions such
> > as strikeForDelta, since it basically applies to vanilla
> > option deltas only.
> check for the payoff and throw, or if you go for the function approach
> use the appropriate vanilla payoff signature
>
> > I can do Vanna-Volga and Malz, if needed.
> It would be appreciated; take a look at Sabr and/or Abcd interpolation
>
> > Also, I'd need some guidance how
> > to start with the contribution...
> Feel fre to ask here on the public mailing list any further questions
>
> ciao -- Nando
>
> > 2008/11/13 Ferdinando Ametrano <na...@am...>
> >>
> >> Hi Dimitri
> >>
> >> > I'm new to this list. I'm thinking about adding new functions
> >> > to the blackcalculator, which are in particular needed in the
> >> > FX area. In particular, I'd like to add:
> >> >
> >> > - premium adjusted black scholes deltas (spot and forward)
> >> > - functions, which return the strike for a given black scholes delta
> or
> >> > ATM
> >> > quotation
> >>
> >> this would be very much appreciated! I've been thinking about tackling
> >> the last issue it's long time now, but never had the occasion.
> >> BTW I've noticed that an algorithm is provided in the last version of
> >> Haug Fomulas book, anyway i don't know if it is the most efficient
> >> algorithm available
> >>
> >> As for premium adjusted black scholes deltas I'm not familiar with FX,
> >> but just go ahead and document the features.
> >>
> >> And while there, anyone willing to provide Vanna-Volga interpolation?
> >>
> >> ciao -- Nando
> >
> >
>