This is brute force version of DDN VRP strategy. The general principles of the strategy remain the same:
- Compare the historical volatility of SPY with the VIX index. The resulting difference (VRP) will represent the premium paid by option buyers for hedging.
- Average the VRP value over several days to smooth the signal.
The subject of optimization here is the window for calculating historical volatility (t1) and the window for averaging VRP (t2).
All combinations of t1 = 1..15 and t2 = 1..12 were analyzed, and the best parameter options were selected. For t1, it is 4 days, and for t2, 8 days.
But calculating volatility based on 4 data points is an absolutely foolish idea. The standard error of the sample standard deviation calculated using the formula:
s ~ SD/(SQRT(2*(N-1)), , where N is the number of points. Let’s say the obtained volatility (SD) is 20%. Then, the standard error on four points would be 8%. Using such a metric for trading is unacceptable.
Since this strategy almost duplicates the DDN VRP, we have excluded it from the test.
Strategy rules
Calculate VRP = VIX – (4days volatility SPY)*100. Average VRP over 8 days. Short VXX if averaged VRP >0, long VXX if averaged VRP < 0.
Strategy Performance
Test period: 2010 – 15 Dec 2023. Costs (brokerage commissions, slippage and borrow costs) are not included.
| Averaged Strategy | Benchmark: Short VXX | Benchmark: SPY | |
| Full Return | 3 500% | 5 850% | 549% |
| Annualized return | 29% | 34% | 12.95% |
| Max DD | -86% | -92% | -34% |
| Sharpe ratio | 0.37 | 0.42 | 0.70 |
