Volatility forecasting
PlannedGARCH-family models against a realized-volatility baseline, evaluated with expanding-window time-series cross-validation.
Research in progress — results, charts, and the risk-metrics table will be published here once the backtest is complete. No performance numbers below are placeholders; this page only carries what's already decided.
Motivation
Volatility forecasts sit underneath position sizing and risk management for almost every systematic strategy — worth getting the evaluation methodology right, not just the model.
Data
Daily returns for a liquid index or ETF, with a long enough history for multiple volatility regimes.
Methodology
GARCH(1,1) and EGARCH forecasts compared against a simple realized-volatility baseline.
Backtest design
Expanding-window time-series cross-validation (never a random shuffle split) — each fold trains only on the past.
Robustness checks
Out-of-sample RMSE and QLIKE against the naive baseline, checked across different volatility regimes.
Limitations
To be written up alongside the results.