GQNR: Combining GARCH Volatility Forecasts with Quantile Regression

Summary

The document introduces a trading model that combines GARCH forecasts of next-period volatility with quantile regression to estimate upper and lower value-at-risk boundaries. It explains how ARCH models use past squared returns and how GARCH adds lagged variance, allowing volatility estimates to reflect persistence. It then describes quantile regression’s asymmetric absolute-error objective and proposes fitting VaR as a nonlinear function of volatility, with polynomial terms and an optimization method such as a genetic algorithm.

The proposed trading interpretation treats boundary breaches as signals for possible short-term reversals: an upper breach may precede a pullback, while a lower breach may precede a rise. The text offers conceptual formulas and trading guidance, but no empirical results or detailed implementation parameters. It identifies model form, optimizer choice, parameter selection, and market randomness as challenges. Suggested mitigations include shorter training windows, smaller margin per trade, trend confirmation from dual moving averages, and a second threshold check. The claims of predictive usefulness are not supported with reported tests, and the approach may be vulnerable to changing market behavior.

Key ideas

• GARCH extends ARCH by including lagged conditional variance as well as past squared returns.
• Quantile regression estimates conditional distribution boundaries through an asymmetric absolute-error objective.
• GQNR models VaR as a nonlinear function of forecast volatility and fits its parameters through optimization.
• The proposed interpretation treats upper and lower VaR breaches as potential short-term reversal signals.
• The document recommends shorter learning periods, smaller trade margin, moving-average confirmation, and a second threshold check.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.