Most American option pricing models use a lattice or partial differential equation (PDE) approach. While these methods are very flexible, they can be relatively slow for applications that require high throughput for large portfolios.
SciComp’s High Speed American Option Pricing Model utilizes an integral equation method for pricing american options. The model is capable of pricing 50,000 contracts per second per CPU for nominal accuracies, e.g. relative PV errors of a few times 10-4. The method is highly parallelizable as well, achieving nearly linear speed-up on multi-core machines, e.g. over 250,000 contracts per second on a six core PC. The High Speed American Option Pricing Model values American options on futures, equity indices, and both non-dividend and dividend paying stocks.
Like all SciComp ready-to-use solutions the High Speed American Option Pricing Model can be enhanced or modified to meet particular modeling needs and is available as C/C++ source code, a Windows/Linux executable, or an Excel spreadsheet and add-in.
Plotted is model throughput in thousands of contracts per second* vs RMS relative PV error over a portfolio of approximately 4,500 options ranging in expiration from one month to one year, and wide ranges of moneyness, volatility, interest rate and dividend yield.
* VS2015 Update 3, 64-bit, Intel i5-8400, Win 10 Enterprise.