Heston Stochastic Volatility Calibration Engine
Options pricing and volatility calibration in C++.
- My contribution
- Implemented FFT pricing, global/local calibration, and an independent PDE pricing method.
C++20 / FFTW3 / Python / Plotly
Explore a volatility surface.
Inspect how a toy volatility surface changes with strike and time to expiry.
Static slice at one year. No market data or calibrated output.
Function, assumptions & limitations
This is a mathematical illustration of curvature, skew, and term structure. It is not the Heston model, a calibrated surface, a forecast, or market data.
σ(k, T) = 0.18 + 0.28(k − 1)² + 0.035√T − 0.045(k − 1)e⁻ᵀ
k is strike divided by spot; T is years to expiry; σ is decimal volatility. Domain: k = 0.70 to 1.30, T = 0.1 to 2.0. There is no arbitrage constraint, dataset, fitting procedure, or empirical validation.
Read the actual Heston pricing & calibration implementation ↗
The problem
Constant volatility assumptions miss the shape of an implied volatility smile. This engine explores a stochastic volatility alternative.
Implementation
Carr Madan FFT pricing feeds Differential Evolution global search followed by Levenberg Marquardt refinement. A Craig Sneyd ADI solver provides a second pricing method. The dashboard supports synthetic prices and parameter recovery.
Evaluation & limitations
Synthetic recovery is distinct from calibration on observed market quotes. Reproducible accuracy and speed benchmarks are not yet available.