Yes, except what is observed (in the US anyway) are Amer-style prices. In the context of a specific model, direct calibration of parameters to those Amer-styleprices would be extremely computationally expensive. So, your method would workone way, but I'm not sure how it could work effectively in the reverse direction, which is needed.I haven't really thought through what is needed here. All I have is the general observation thatin specific cases (like the Heston model, for example), the "basic method" that you first postedoften doesn't work very well.

I can confirm the performance issue, I've actually used the inversion-oriented (not even the calibration-oriented) method in my project (options on futures, American-to-European conversion), and the performance was _vastly_ different (under 1 day vs. not-done-in-14-days) depending on which American-option-pricing-engine was chosen, just as the quality (not just the approximation error, but also the percentage of prices where the root-finding failed to converge at all or got stuck at the boundary).I think that's a very interesting feature request (would've loved to have had that back then instead of experimenting with the engines one-by-one myself).I also don't think that we can expect performance wonders from the model-dependent-calibration-oriented method (other than deployment in a massively environment), at least I have to humbly admit it's neither trivial nor obvious to me right now.Let us know if you have anything further on this (or other suggestions, of course)!

Yes, direct calibration to American prices is very expensive. One possible way to do it is to use adjoint approach, to speed up the computation of the gradient. An example (although not exactly what I have in mind) was done in work by Maruhn et al, 2009, Adjoint-based Monte Carlo calibration of financial market models. Their example is based on writing adjoint equation, and then discretize. What I have in mind is to write adjoint of the code directly

Hello, i dont have an idea rather a question.is the scope already defined?e.g.- what types of securities are covered?- is it just about derivatives?- what is about bonds/loans with embedded options?

Quote- what types of securities are covered?- is it just about derivatives?- what is about bonds/loans with embedded options?I think the group should at least form some vision regarding these issues. More than equities? Are 'librares' the way to start? To answer hamster; the scope not defined, although it should be asap.

Shouldn't scope and project structure encourage both bottom-up/push and top-down/pull contributions? If someone has ready-made code or things that they really want to do, then they should be able to voluntarily do it and push it out to the project community. Yet, it's probably also good to maintain a "centralized task-list which is part of a greater plan" to help pull the most valuable contributions possible. That way, others can see where they can make the greatest contribution if they don't have anything specific already in mind.I can also imagine that some contributors might find the "integrated, documented, tested, and peer-reviewed" criteria to be too daunting. They might be willing to donate "as-is" code that others might integrate, document, test, etc. A centralized task list lets everyone see the gaps (e.g., financial instruments with no models, untested code, etc. ), determine if there is demand for a contribution, and to contribute in the way they can.

QuoteOriginally posted by: Traden4AlphaShouldn't scope and project structure encourage both bottom-up/push and top-down/pull contributions? If someone has ready-made code or things that they really want to do, then they should be able to voluntarily do it and push it out to the project community. Yet, it's probably also good to maintain a "centralized task-list which is part of a greater plan" to help pull the most valuable contributions possible. That way, others can see where they can make the greatest contribution if they don't have anything specific already in mind.Indeed. A good balance between content and administation is vital. We do not want to stifle new ideas.,

It seems like the project needs some kind of the roadmap/status tracker and content management system that lets people contribute new research papers (or links to the literature), Matlab code, etc. and others can see the progressions and linkages from math to sandbox code to released code.Has any open source system ever used social media techniques to encourage semi-directed participation? I can envision a system where everyone votes on project priorities and the contributors that address those priorities get points or virtual currency in proportion to the priority of the need they filled. If a person's votes are weighted by their virtual currency earnings, then those that make the greatest contributions to the most crucial needs of the project will have the largest say in the direction of the project. The most valuable contributors could then strongly encourage others to contribute.