VXX holds short-term VIX futures and loses value most of the time as it rolls them. After a sharp fall, the bet is that the decline partly reverts. Naked calls on a volatility product can lose everything on a spike, so the idea was to express the view with call spreads, where the maximum loss is fixed in advance.
volatility mean reversion
VXX call spreads
The idea
How it was tested
Pre-registered on 2026-08-31, as a variant of a VXX mean-reversion sleeve. The signal: VXX down more than 15% over 20 days and RSI(14) below 40. The structure: a call spread 5% wide, 30-45 days to expiry, bull when VIX is below 15 and bear when VIX is above 30. Exit at 50% of max credit, after 30 days, or at a loss equal to the credit.
Where it died
The backtest ran the same day and produced zero trades. The factor library it read had no VIX history, so the VIX conditions could never be evaluated. There were no historical VXX option chains either, so even a working signal would have had no prices to trade. It was archived on 2026-08-31 without a single measurement, and the verdict is void.
The record
Thesis
After VXX has fallen sharply (20-day return below −15% and RSI below 40), its decline tends to revert. Express that with call spreads instead of naked calls, so a volatility spike has a defined maximum loss.
Specification
- signal
- VXX 20-day return < −15% and RSI(14) < 40
- structure
- call spread 5% wide, 30-45 days to expiry; bull spread when VIX < 15, bear spread when VIX > 30
- exit
- 50% of max credit, 30 days, or loss equal to the credit
Gates
Pre-registrationpass
Was the hypothesis, its parameters and its pass/fail rule written down before any data was looked at?
Gate 7 — Jensen alphanot evaluated
After regressing out market exposure, is there alpha left?
The backtest produced 0 trades.
Why it died
The backtest produced zero trades because the factor library had no VIX history, and the lab has no historical VXX option chains either. It was archived with no measurement of any kind.
What it taught
Check that every input the signal reads exists in the data before pre-registering. A zero-trade run is a data audit, not a verdict.
Verdict history
- 2026-08-31Pre-registered as a variant of a VXX mean-reversion sleeve
- 2026-08-31Backtest: 0 trades (no VIX history in the library); archived Void