Help us double down on what's working, instead of guessing. Takes 5 seconds, totally optional.
research · 3 min read
What Predicted the COVID Crash? We Tested the Signals Option Sellers Use
Nothing predicted the February 2020 crash. We replayed dealer gamma, the volatility risk premium, the VIX term structure and three volatility models on 11 symbols. Two signals failed on opposite halves of the fall, and a rule that needed both sold nothing until the low.
Short answer: nothing predicted it. At the 21 February 2020 close, every volatility forecast we tested put SPY's volatility for the next month between 12% and 14%. It came in at 76%. What helped option sellers was a pair of signals that failed on opposite halves of the crash. Before the drop, the volatility risk premium was negative, so a VRP rule saw no premium worth selling. During the drop, dealer gamma was negative, so a gamma rule stopped selling. A rule that needed both sold nothing from 3 February to the 23 March low.
76%SPY volatility in the 21 sessions after 21 February. Forecasts said 12% to 14%.
0SPY sessions from 3 February to the 23 March low on which the gamma-and-VRP rule sold
0%of the sell-every-day crash loss kept by that rule (16-delta 1-month put, median of the 10 symbols that lost money selling every day). Gamma alone kept 67%, VRP alone 26%.
6 of 6selloffs since 2018 where requiring both kept less of the loss than either rule
Alpha members
This section is for Alpha members
Alpha members get the full method and code behind this study, and the Historical API to run it themselves.
No. At the 21 February 2020 close, GARCH, GJR-GARCH, HAR and implied volatility put SPY's volatility for the next 21 sessions between 12% and 14%. Realized volatility over those sessions was 76%. No signal we tested warned of the size or timing of the fall.
Did dealer gamma (GEX) warn before the crash?
Only just. SPY dealer gamma turned negative at the 21 February close, after a 1.0% drop and one session before the 3.3% drop on 24 February. Negative gamma is common, though: the gamma rule allowed selling on only 36% of SPY sessions from 2018 to 2026. It worked as a switch that turned off once the fall began, not as a forecast.
Why did the volatility risk premium rule avoid the start of the crash?
In early February 2020, implied volatility was below realized volatility, so the volatility risk premium (VRP) was negative and the rule saw no premium worth selling. It allowed selling on 21% of SPY sessions from 3 to 21 February. Once the fall started, implied volatility jumped above realized and the rule sold on 90% of sessions.
What does requiring both gamma and VRP do?
The two failed on opposite halves of the crash, so a rule needing both sold on no SPY session from 3 February to 23 March 2020. For a 16-delta 1-month short put it kept 0% of the loss of selling every day (median of the 10 symbols where selling every day lost money). It also kept less of the loss than either rule alone in all 6 selloffs we checked since 2018. Its Sharpe ratio from 2018 to 2026 was 0.41, against 0.49 for selling every day.
Should you close short options when dealer gamma turns negative?
It depends on the trade. Closing on negative gamma cut the crash loss of a 16-delta 1-month put to 11% of selling every day, but over 2018 to 2026 its Sharpe ratio fell from 0.49 to 0.08. For the 16-delta 1-week put it rose from 0.47 to 0.61. Gamma turns negative often in normal markets, and longer trades closed at poor prices too many times.
