Graveyard / Options and volatility / US stocks
Variance risk premium timing
Does a high gap between implied and realised volatility predict strong stock returns?
The idea
Implied volatility usually exceeds the volatility that follows. Academic work (2009) found that when this gap is wide, the S&P 500 tends to earn high returns over the next one to four months. Hold SPY when the gap is high, stay out otherwise.
How we tested it
| Data | Cboe VIX history and SPY. The gap is VIX squared over 12, minus the sum of squared SPY daily log returns over 21 sessions. Long SPY while the gap is at or above its trailing 252-session median, traded at the next closing auction. A second test did the same per asset over 11 ETF and volatility-index pairs. |
|---|---|
| Period | February 1994 to October 2026, split at 27 March 2020. |
| Costs | 1 bp per side; stress case 6.5 bp. |
| Holdout | In-sample and out-of-sample Sharpe, plus a filter test. |
| Control | SPY buy and hold over the same window (Sharpe 0.65). |
Result
| Trades (primary) | 556 |
|---|---|
| Sharpe full / in-sample / out-of-sample | 0.52 / 0.37 / 1.18 |
| Buy-and-hold SPY Sharpe | 0.65 |
| Max drawdown / profit factor | −43.6% / 1.62 |
| Sharpe at stress cost | 0.38 |
| Sharpe in bear / bull / chop markets | −0.78 / 2.29 / 0.01 |
| 11-pair version, Sharpe full / in / out | 0.59 / 0.46 / 1.15 |
| 11-pair version, legs worse than buy and hold | 10 of 11 |
Why it failed
In this daily-data form the signal adds nothing over buying and holding: the full-period Sharpe of 0.52 is below the 0.65 of simply holding SPY, and the drawdown is still −44%. An extra volatility filter lowered the out-of-sample Sharpe (0.99 against 1.18) instead of raising it, which breaks the rule that a filter must help.
What would make it worth retesting
No reopen condition is stated.
Past results of a backtest say nothing certain about the future.