VIX Adaptive Momentum Adaptations
Andrew Miller, VIX and Trend-Following, the Killer Combo? (Alpha Architect original firm research, September 28, 2017), chooses 10/3/1-month momentum from VIX states and replaces each negative selected slot with T-bills. Our native Top 1 maps the indices to SPY/VXF/EFA/AGG: VIX SMA(40) <= 18 uses 210-row returns; otherwise SMA(20) < 32 uses 63 rows; otherwise 21 rows. We assign equality at 32 to the high state. Missing VIX windows select BIL. Only strictly positive scores qualify; zero is excluded. Top 2 renormalizes positive survivors (one survivor gets 100%), instead of retaining a half-cash slot; all-negative scores select BIL. These ETF, window and allocation changes are material adaptations. The author's results were mixed and do not establish a reliable future edge. Native total-return ETF history, monthly scheduling and next-session signal lag apply; available history bounds the backtest. Taxes and trading costs use the normal request settings (zero in the gallery). No synthetic pre-inception history or source-backtest parity is claimed. Source research does not validate this adaptation or future returns.
- Type: Tactical
- Frequency: monthly
- Asset classes: US Equity, International Equity, Bonds, T-bills
- Backtest window: 2007-05-30 to 2026-10-09
- 1Y
- 13.0%
- 3Y
- 50.0%
- CAGR
- 12.9%
- Max DD
- -22.5%
- Sharpe
- 0.57
Methodology
- Select up to 1 positive-return ETFs using 210/63/21 rows according to VIX SMA states; equal-weight survivors and use BIL if none qualify.
- SMA(40) <= 18 is calm; otherwise SMA(20) < 32 is elevated; equality at 32 is high.
- Native total-return ETF history, monthly scheduling and next-session signal lag apply; available history bounds the backtest. Taxes and trading costs use the normal request settings (zero in the gallery). No synthetic pre-inception history or source-backtest parity is claimed. Source research does not validate this adaptation or future returns.
Headline backtest performance as of 2026-10-08.