Wonk Post: Where Do Your VIX Strategy Returns Come From?

Something I find fascinating is comparing the source of returns for different VIX ETP strategies:

  • A strategy might generate returns by predicting changes in the overall level of VIX futures, most likely because of changes in the VIX index, either for example because the VIX is overbought or oversold (ex. TM’s RSI(2)) or because the VIX is selling at a premium or discount to historical volatility (ex. DDN’s VRP).
  • Alternatively, it might predict changes in VIX futures, regardless of what’s happening with the VIX index, either for example by measuring the premium between futures and the spot index (ex. VIX vs Front Month Futures) or between distant and nearer contract months (ex. First vs Second Month Futures).

Those two bullet points might seem similar, but in fact they’re very different sources of return. The next two graphs test for the first bullet point: changes in the VIX index (graph #1) and changes in the general level of VIX futures (via the 30-day constant maturity price of VIX futures, graph #2).

I’m using the VIX vs Front Month Futures strategy as our sample strategy (click for strategy rules).

20140616.VIX-Index

20140616.30D-CM

The blue line on each graph assumes that we bought either the VIX index (graph #1) or the 30-day constant maturity price of VIX futures (graph #2) when our sample strategy was short the VIX (i.e. long XIV or ZIV). For simplicity’s sake, I’m ignoring long VIX trades.

Of course, we can’t actually buy the VIX index or 30-day constant maturity, but that doesn’t matter here; this is only a proof of concept.

In a perfect world, we would like to see each line going down, indicating both the VIX and the overall level of futures fell when the strategy was short the VIX.

In fact, we see something very different. The VIX index consistently rises when the strategy is short (which is bad), but those increases are washed out of the 30-day constant maturity price (which is good).

What’s going on here?

Recall that this strategy is short the VIX (long XIV or ZIV) when the front month futures contract is selling at a premium to the spot index. That scenario is more likely to happen when the VIX is low, meaning that a rising VIX is more likely than one that continues lower. That explains why there tends to be an increase in the VIX index.

But that increase doesn’t carry over to the VIX futures because futures traders aren’t (on average) following that increase in the spot index. There are lots of examples of VIX futures not following suit with changes in the VIX, the quintessential one being the weekend effect.

* * *

So what? So far, these are just interesting observations. What about actual VIX futures and ETPs, i.e. the things that we can actually trade.

The next two graphs assume we bought either the front month VIX futures contract (graph #3) or the VIX ETN VXX (graph #4) when our sample strategy was short the VIX.

20140616.Front-Month

20140616.VXX

In graph #3 and 4 we see that, despite the strategy’s lack of connection to the overall level of VIX futures (i.e. chart #2, 30-day constant maturity), it has been successful capturing the VRP that exists between the front month and spot VIX (graph #3) and between the second, front, and spot VIX (graph #4).

And that brings us to our last chart, which assumes our return (when short) equaled the daily change of the VXX minus the daily change in the 30-day constant maturity price.

Here we’re isolating just the VRP between the second, front and spot VIX, without those annoying changes in the overall level of VIX futures to get in the way (because we know that the strategy is neutral to the overall level of futures, see graph #2).

20140616.VXX-30D

Why perform this type of breakdown? Why not just look at standard performance metrics, like Sharpe Ratio, UPI, etc. when judging a strategy?

Because, understanding how the sources of return for various strategies differs is key when combining strategies in a way that they are most likely to complement each other out-of-sample.

Traditional methods like correlation matrices are too simple by half, especially because (a) we have so little historical data to consider, and (b) the volatile nature of VIX ETPs means that a handful of events (like the 2008/09 crash) have an outsized impact on backtested performance.

Click to see Volatility Made Simple’s own elegant solution to the VIX ETP puzzle.

Good Trading,
Volatility Made Simple

Posted in Strategy Mechanics.