Australian institutions often face markets where volatility regimes shift abruptly. To respond, calibration begins by gathering historical data, including prices and implied volatilities. The initial parameter estimates set the baseline for fitting the stochastic volatility model. Our team uses optimization routines—such as least squares or maximum likelihood—to refine these values. Each step, from collecting relevant data to selecting the right error metric, impacts the model’s sensitivity and predictive reliability. Historical case studies, like the post-2015 commodity price swings, show that even small changes in calibration can lead to large differences in daily pricing outcomes. Calibration is rarely a one-off exercise. Instead, it’s an iterative process, revisited whenever market conditions or data quality change. By repeating calibration frequently and monitoring results, practitioners ensure models remain effective as market dynamics evolve.

A successful calibration is stress tested using out-of-sample scenarios. This means running the calibrated model against market events not used in the fitting process. For example, periods of heightened volatility—such as those triggered by unexpected regulatory announcements—can reveal if the model’s parameters are robust or prone to overfitting. Our team tracks deviations between predicted and actual volatility surfaces, adjusting parameters where persistent gaps arise. Case studies from recent years highlight the need to balance model complexity with interpretability. Overly complex models might capture more detail but are harder to maintain and explain. Simpler models, while easier to calibrate, may miss important features. The key is to choose an approach that aligns with portfolio objectives and available data, revisiting calibration as new scenarios unfold.