SciComp releases SciCal calibration software for derivatives pricing models
SciCal specifically calibrates models that use stochastic volatility (SV) and stochastic volatility + jumps (SVJ). SciCal was developed in response to the growing use of SV and SVJ models to price structures that are sensitive to forward volatility skew. SciCal can calibrate from a handful to hundreds of options to market data in under a minute.
"Like other calibrators, SciCal searches for parameters that minimize the model-to-market mismatch. What makes SciCal different is the way it searches and how fast it is accomplished," said Curt Randall, Executive Vice President for SciComp. Standard calibration algorithms start from an initial 'best' guess for model parameters and run until reaching a minimum. Unfortunately, there are typically many local minima, and the initial guess may not lead to the global minimum. Small moves in the market may cause a jump from one minimum to another, leading to instability. SciCal performs a stochastic search of the entire parameter space, climbing out of local minima to find the global minimum.
"Finding the true global minima is the whole point of calibration. If your calibration has unstable parameters, model risk can increase as the market moves," added Randall.
Aimed at quantitative analysts, traders and risk managers can benefit from SciCal runs in Excel or as a standalone executable for Windows or Unix. SciCal allows expert users to adjust important search parameters for complete control over the calibration process, or SciCal can make optimal choices for the user.
This article appeared in Wilmott Magazine.