Volatility ETFs
The VIX curve is in contango, so the one strategy that passed our test, holding SVXY, is on. Every long-volatility ETF lost money when held.
- SVXY: hold. Sell at the next open after VIX ÷ VIX3M closes at or above 1.00
- VXZ or VIXM: the hedge you can hold for weeks (it decays about 16% a year)
- VIXY or UVXY: a hedge for days, against a specific event
- Don't hold VIXY, VXX, UVXY or UVIX: -48% to -79% a year
- Don't hold SVIX: the full-size short lost -96% in one day in 2018
- Don't act a day late: the delay erased most of the edge
13.4 years of growth, $1 at the start
Rebuilt from official VIX futures settlements at each product's current leverage and fee. Log scale: each gridline is ten times the one below.
What each product is, and what holding it did
All track the same VIX futures. They differ in direction, size, and how far out on the curve they sit, and that decides everything.
| Ticker | What it is | Exposure | Fee | Held for 13.4 yearsRebuilt from VIX futuresEach product rebuilt from official Cboe futures settlements at its current leverage and fee, so every ticker covers the same 13.4 years, 2013 to 2026. Compound return per year if you simply held it. | Worst day | Real, since launch |
|---|---|---|---|---|---|---|
| SVXY | Short vol, half size | -0.5x 1-month | 0.95% | +10.6% a year | -48.1% | +514% since 2011 |
| SVIX | Short vol, full size | -1x 1-month | 1.35% | -7.0% a year | -96.1% | +99% since 2022 |
| VIXY | Long vol | +1x 1-month | 0.85% | -47.6% a year | -26.0% | -100.00% since 2011 |
| VXX | Long vol | +1x 1-month | 0.89% | -47.6% a year | -26.0% | -99.04% since 2018 |
| UVXY | Long vol, 1.5x | +1.5x 1-month | 0.95% | -68.3% a year | -38.9% | -100.00% since 2011 |
| UVIX | Long vol, 2x | +2x 1-month | 2.19% | -78.7% a year | -51.9% | -99.98% since 2022 |
| VXZ | Long vol, mid-term | +1x 5-month | 0.89% | -15.5% a year | -15.5% | -34% since 2018 |
| VIXM | Long vol, mid-term | +1x 5-month | 0.85% | -15.5% a year | -15.5% | -96% since 2011 |
SVXY was −1x until Feb 27, 2018 and UVXY was 2x; both were cut after the XIV collapse, so the real since-launch figures mix two products. The 13.4-year column applies today's terms throughout.
Every strategy we tested
Rules fixed before the test, 0.10% cost per switch, results split before and after 2020. Strategies that failed are listed too.
When you act decides whether it works
The same contango rule on real ETF prices, varying only when the switch happens. SVXY since March 2018, after it became −0.5x.
| SVXY, real prices | Per year | Worst fall | Worst day |
|---|---|---|---|
| Buy and hold | +12.3% | -62% | -21% |
| Switch at the same close | +14.2% | -36% | -17% |
| Switch at the next open | +14.4% | -44% | -22% |
| Switch one full day late | +7.0% | -59% | -21% |
| SVIX, real prices | Per year | Worst fall | Worst day |
|---|---|---|---|
| Buy and hold | +16.7% | -79% | -39% |
| Switch at the same close | +26.8% | -50% | -22% |
| Switch at the next open | +8.9% | -75% | -40% |
| Switch one full day late | -4.5% | -85% | -39% |
SVIX (−1x, since 2022) shows why size matters: switching at the next open earned +8.9% a year against +26.8% at the same close, because a full-size short loses the most in the overnight gap. On SVXY the open matched the close on return, with a deeper worst fall.
Five shocks
Return over each crash window. Exited at the close, the contango rule never lost more than 12%. One day late, it did worse than holding in 2 of the 5.
| Shock window | Aug 2015 | Feb 2018 | Mar 2020 | Aug 2024 | Apr 2025 |
|---|---|---|---|---|---|
| SVXY, held | -26% | -48% | -54% | -17% | -26% |
| SVXY with the contango exit | -7% | -10% | -12% | -9% | +1% |
| …the same exit, one day late | -13% | -53% | -16% | -33% | -13% |
| SVIX, held | -48% | -96% | -81% | -38% | -48% |
| VIXY, held | +71% | +76% | +253% | +15% | +55% |
| UVXY, held | +117% | +96% | +485% | +13% | +80% |
How this was tested, and its limits
Data. The two S&P VIX futures indices every one of these products tracks (short-term: 1-month, rolled daily from the first to the second future; mid-term: 5-month, the fourth to seventh) were rebuilt from official Cboe settlements, May 2013 to Oct 2026 (3,372 trading days). Daily returns track the real ETFs closely: correlation 0.98 for SVXY, 0.99 for UVXY, 0.95 for VIXY. The rebuild ignores the T-bill interest on collateral, so it understates SVXY by about 1.6% a year.
Rules. Written down before the test: contango below 1.00, the Simon and Campasano ±0.10 basis, and Cooper's premium switch as published. The test we fixed in advance switched at the close one day after the signal, and there the contango rule rated only Weak. It is Robust when switched at the signal's own close. The next-open check on real prices was added after we saw that. 0.10% of notional per switch; cash earns nothing. Robust means the Sharpe's by-week bootstrap range stays above zero, both eras are positive, and the worst fall is smaller than holding. Two independent checks re-computed every headline result from scratch and audited the index rebuild.
Fragility. Much of the gap between acting promptly and a day late comes from a handful of days (Feb 5, 2018, Aug 5, 2024 and Apr 3, 2025). Its drawdown protection held at every exit line from 0.90 to 1.01, but it beat holding on return only for lines from 1.00 up. Treat it as crash insurance that pays for itself, not as a return booster.
Timing. VIX and VIX3M print their official close at 4:15 pm New York time, 15 minutes after the ETFs close, so a same-close switch needs a near-close reading. This site updates before the next open, which is why the next-open result is the one to plan on.
Sources. Simon and Campasano, "The VIX Futures Basis: Evidence and Trading Strategies" (2014); Cooper, "Easy Volatility Investing" (2013); ProShares on the Feb 27, 2018 leverage change. Full method and every number: research/ETF_EVIDENCE.md in the source.