SciFinance, the premier derivatives pricing code generation technology, continues to evolve with more powerful functionality.
By automatically translating model specifications for any financial derivative in
to fully documented C++ or CUDA source code, SciFinance provides an optimal tool for model
development, prototyping, model validation, risk management and quantifying counterparty exposure.
SciFinance does not impose a set of pre-implemented, "black-box" canned models, but instead allows users to easily and rapidly create bespoke models, thus facilitating the evaluation of a model's conceptual soundness and its sensitivity to changes in model parameters.
A new library for complex basket equity linked notes (ELNs).
ELNs often have complex features, e.g., knock-in of an alternative payout, coupons based on equity performance including range accrual that may knock-in or knock-out and automatic or discretionary early redemption. SciFinance now includes a collection of ready-to-use Monte Carlo macros that capture these complex, but commonly found ELN features.
Using the macro library even the most complex ELNs can be specified in a half page. SciFinance then generates the corresponding pricing model code. For a GPU-enabled version of the code, with identical numerical results, simply add a single keyword to the specification and resynthesize the code.
PDE and MC methods for
Overhedges are useful when managing the exposure of a financial product that contains discontinuities in payoff, coupons, or other cashflows (as in ELNs), by limiting the frequency and cost of hedge rebalancing. SciFinance includes several methods for easily modelling over/under hedges for both PDE and Monte Carlo based pricing models (including GPU Monte Carlo code).
Bilateral counterparty risk and funding costs
in derivative pricing can often be represented as nonlinear terms in PDE models. SciFinance, which includes several choices for nonlinear PDE solutions, therefore allows these important components of XVA to be modelled by robust and computationally efficient PDE methods.