A high-frequency approach to Realized Risk Measures

Battle Count: 79

Highly relevant for quantitative trading as it provides a novel method for estimating tail risk measures (VaR and ES) from high-frequency data, which is directly applicable to position sizing, portfolio risk management, and regulatory compliance. The method leverages minute-by-minute data commonly available to quantitative traders. The out-of-sample forecasting framework with AR(1) models is directly applicable to trading risk management. However, the method is more focused on risk estimation than alpha generation.

Moderate complexity. The pipeline involves multiple steps: subordinator construction, MA(1) filtering, distribution fitting via MLE, and scaling via either characteristic function (requiring numerical integration and root-finding) or Monte Carlo simulation. The characteristic function approach requires careful numerical implementation (Gil-Pelaez inversion, Brent's method). The MA filter adds complexity to the scaling step. However, the code is publicly available, and the algorithms are well-documented with pseudo-code. Computational time is manageable (sub-second per day for MC, up to ~600ms for CF approach).

Code is publicly available on GitHub (https://github.com/fgt996/Realized_Risk_Measures). However, the data used (minute-by-minute equity prices) are not available due to copyright restrictions. The methodology is well-described with algorithms and pseudo-code provided. Parameters and experimental settings are clearly specified. Reproduction would require access to proprietary high-frequency financial data.