Graveyard / Options and volatility / US stocks
Short volatility: VIX ETF carry
Can you earn the VIX futures roll by being short volatility?
The idea
VIX futures usually trade above the spot VIX, and roll down toward it as time passes. A short-volatility ETF (or a short position in a long-volatility ETF) collects that roll as well as the premium for insuring the market. Stay in while the curve is upward sloping, leave when it inverts.
How we tested it
| Data | VIX and VIX3M daily closes, VIXY and SVXY daily prices. Long a synthetic −0.5x VIXY when the VIX/VIX3M ratio is at or below the entry level, exit at the next open when it reaches the exit level or a stop is hit. Grid: entry 0.85, 0.90, 0.95; exit 0.95, 1.00, 1.05; stop none, 10%, 20%; with or without a volatility filter (48 configurations). A second test shorted VIXY on the first session of each week from 09:35 to the close. |
|---|---|
| Period | January 2016 to October 2026 (includes February 2018, March 2020, 2022, August 2024 and April 2025). The weekly test has 562 sessions. |
| Costs | 0.13% round trip for the carry test. Weekly test: 5 bp per side plus 3% a year borrow. |
| Holdout | Full, in-sample and out-of-sample gate. The drawdown limit was 25%. |
| Control | No random-entry control stated. |
Result
| Configurations passing | 0 of 48 |
|---|---|
| Total return of the configurations | +60% to +320% |
| Max drawdown of every configuration | −31% to −62% |
| Full-period Sharpe | 0.22 to 0.60 |
| Out-of-sample Sharpe / profit factor of the in-sample pick | −0.13 / 0.88 |
| Trades per configuration | 10 to 53 |
| Weekly short-VIXY test: Sharpe / profit factor | 0.39 / 1.17 |
| Weekly short-VIXY max drawdown | −45.4% (−39% in 2018) |
Why it failed
The carry is real but the tail is the problem. The ratio only inverts after the first day of a volatility spike, and that first day is where the loss happens, so the exit is one day late for gap shocks. Every configuration lost more than 25% from peak at some point. The weekly short-VIXY rule earned +52.5% over the period but with a −45% drawdown.
What would make it worth retesting
Only with a leading tail signal that the ratio does not give, such as intraday spike detection that exits the same day, or an options-based hedge, and only if drawdown can be brought under 25%. Not by re-sweeping the ratio thresholds.
Past results of a backtest say nothing certain about the future.