|
|
I've gotten pricing engines written for all single period pricers, and
I'm about to start with the multiperiod options. There is an
architectural issue that I'm not quite sure how to deal with.
The class structure for the engines that I've written so far look like
european -> fdvanillaengine -> VanillaOption::Engine -> GeneralEngine
<VanillaOption::arguments, VanillaOption::results>
I'd like have the multi period engines use most of the code in fdvanilla
engine. The trouble is that the multi period engines will need
start/stop date information that is in DividendVanillaOption::arguments
and hence inaccessible if I make it a subclass of fdvanilla engine.
Thoughts?
Also any input will be appreciated on
1) how to model the convertible bond class in C++, and
2) anyone who is interested in adaptive binomial trees? I've read
Figlewski and Gao's paper on using adaptive meshes to model American
options in the Journal of Financial Economics 53 (1999) 313-351, but was
wondering if there have been any more recent additions to it or if
anyone out there is working on this problem. The issue that I have with
the Figlewski paper is that the method that he describes is "irregular"
but it isn't adaptive in the sense that a truly adaptive mesh would
"autozone" itself. This is something I might be able to contribute
usefully to since my dissertation was in large part on adaptive meshes
to solve PDE problems and it would also get me into the binomial code.
|
Locked
