Project background
This project was completed as part of the Erdős Institute's Summer 2026 Quant Finance Bootcamp. My group's project conducted a comparison of different market volatility models across different market regimes.
I built code to implement a version of the path-dependent volatility models proposed by Guyon-Lekeufack (2023). In general, the model strongly outperformed a custom ML model built by one of my teammates, as well as the classical HAR and GARCH models.
Baselining against Black-Scholes constant volatility, we compared rolling annualized realized 22-day volatility for these different models.
See the links below for more information and details about our project!
Someday I hope to add a live tool for playing with the models on any training data and tickers.
Our submission materials and certificate
Hosted on the Erdős Institute's website
Contains our code and details on implementation.