From rough to multifractal multidimensional volatility: A multidimensional Log S-fBM model

Battle Count: 80

Highly relevant for quantitative trading as it provides a unified framework for modeling multivariate volatility dynamics that captures both rough (H≈0.15) and multifractal (H≈0) behaviors observed empirically. The model's ability to explain why individual stocks show multifractal behavior while indices show rough behavior is directly applicable to portfolio construction, risk management, and volatility forecasting. The GMM calibration procedure enables practical implementation. However, the model is primarily a volatility model and does not directly generate trading signals.

High complexity due to: (1) multidimensional Gaussian process construction with non-homogeneous white noise, (2) complex cross-covariance structures involving co-Hurst and co-intermittency matrices, (3) small intermittency approximation derivations, (4) GMM calibration requiring pairwise estimation for all asset combinations, (5) simulation requiring spectral-domain circulant embedding methods, (6) constraint handling (positive definiteness, conditional negative definiteness). The mathematical framework is sophisticated but the actual calibration uses standard optimization tools (scipy.optimize, L-BFGS-B).

The paper provides detailed mathematical formulations, simulation procedures (Appendix F), and uses publicly available S&P 500 data from Yahoo Finance. However, no code repository is mentioned. The GMM implementation uses scipy.optimize with L-BFGS-B algorithm and Newey-West HAC estimator. Synthetic data generation procedure is described but not fully reproducible without code.