AR(2)-GARCH Mean-Reversion Signals with Dynamic Volatility Bands

Summary

The strategy combines an AR(2) model, estimated from recent returns using Yule-Walker equations, with a GARCH(1,1) model for conditional volatility. The autoregressive component forecasts returns from the prior two observations, while the volatility model updates its estimate using both recent squared residuals and prior conditional variance. The forecast and volatility estimate define upper and lower bands that adapt to market conditions.

It proposes buying when price falls below the lower band and shorting when price rises above the upper band, with optional RSI confirmation and percentage-based stop-loss and take-profit orders. The document describes stability constraints for the AR coefficients and a convergence constraint for GARCH parameters. It provides model equations and implementation details, but no performance results. Its stated caveats include changing market conditions, dependence on the estimation frequency and window, and trading costs; the parameters and assumptions require out-of-sample evaluation.

Key ideas

• The AR(2) component forecasts returns from the two most recent return observations.
• GARCH(1,1) estimates changing volatility from recent residuals and prior conditional variance.
• The forecast and volatility estimate create dynamic bands used for mean-reversion entries.
• An optional RSI filter and fixed percentage exits supplement the band signals.
• Model stability, parameter variation, data frequency, and trading costs are practical concerns.

Tags

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