This is a test of Trading the Odds’ Volatility Risk Premium (VRP) trading strategy (1). The strategy is very similar to the Brute Force VRP strategy we shared back in July, and compares the VIX spot to historical volatility to predict changes in VIX ETPs like XIV and VXX. Frank, the brain behind Trading the Odds, has been doing some excellent VIX trading research lately and I highly recommend the follow. It’s good to see more quantitative types talking about the subject.
Edit: see footnote re: difference between my results and those shown by Frank on his site.
Below are the results of TTO’s strategy in blue, versus the Brute Force VRP strategy we presented in grey, both trading XIV and VXX from 07/2004 to present. Read about test assumptions, or get help following this strategy.
- At the close, calculate the following: the 5-day average of the [VIX index – (2-day historical volatility of SPY * 100)]. Note that historical volatility is based on the natural log of each day’s % change.
- Go long XIV at the close when the result of the above formula is greater than 1.25, otherwise go long VXX. Hold until a change in position.
Regardless of any difference in the stats, these two strategies are so similar that I would be more or less equally confident in either strategy’s performance in the future. As with most trading strategies, the concept being exploited is much more important than the specific parameters chosen.
The differences between the two strategies follow:
- TTO’s strategy uses a 2-day standard deviation of SPY (as opposed to the Brute Force strategy’s 4-days). That shorter lookback is going to lead to much more volatile results for the standard deviation calculation, which is likely why this strategy trades nearly 3-times as often.
- Rather than using zero as the threshold for buying XIV vs VXX, TTO’s strategy uses what I assume was an optimized value of 1.25. That’s going to make the strategy slightly more likely to go long VIX (ex. long VXX). I don’t have an issue with that, but I might also caution about adding unnecessary extra parameters, unless it led to a large and consistent improvement in performance, as it implicitly increases the risk of curve-fitting results.
Again, as I’ve stressed repeatedly, the concepts being exploited are much more important than the specific parameters chosen. All sets of parameters will, over the long-term, rise or fall together based on the success or failure of the core concept. Here that concept is to compare the VIX spot to historical volatility to predict VIX ETP returns. This the third strategy we’ve covered that is exploiting this observation (the others being DDN’s VRP and our own Brute Force VRP).
Note that I don’t see this category of strategies as a panacea. Recent success hides the fact that for both 2012 and 2013, these strategies trailed buy & hold badly. I think a more holistic approach, that considers not just the concept presented here, but many of the other concepts we talk about here at VMS as well, will be the better play over the long-term.
Having said all of that, I think that there’s value in presenting divergent views, so when this strategy (like most we cover on this blog) signals a new trade, we’ll include an alert on the daily report sent to subscribers. This is completely unrelated to our own strategy’s signal; it just serves to add a little color to the daily report and allows subscribers to see what other quantitative strategies are saying about the market.
Click to see Volatility Made Simple’s own elegant solution to the VIX ETP puzzle.
Volatility Made Simple
- These type of strategies, comparing the VIX spot to historical volatility, have become known as “volatility risk premium” or VRP strategies. In truth, there are multiple VRPs in the VIX complex (VIX spot vs realized volatility, VIX futures vs realized VIX, etc.), so admittedly, the term is probably not the most accurate. In any case, we stick with that convention here.
- There is a pretty significant difference between the results I’ve shown here and those shown on Frank’s site. After some quick analysis, it looks like Frank might have used the S&P 500 cash index (GSPC) as opposed to the non-dividend adjusted SPY. Why would there be such a large disparity in results using two such similar indices? To some degree, it’s likely the result of curve-fitting, especially bearing in mind that optimized 1.25 cutoff. That isn’t intended as an affront, as all backtests implicitly include some amount of curve-fitting. That also doesn’t mean using the cash index is or is not superior, only that the results are probably overly optimistic. In any case, I’ll be posting follow up results in the near future.