Evolution Capital’s VIX Trading Strategies (Part I)

In the coming week I’ll be breaking down the strategies in Evolution Capital’s paper Volatility: A New Return Driver?

In this post I break down strategies #1 and #2 from the paper, which were meant to show the benefit of comparing not just the VIX to front month VIX futures (to judge whether VIX futures are in contango or backwardation), but also front and second month VIX futures.

20140924.01

The graph above shows two strategies from the paper trading the VIX ETN VXX (read about test assumptions):

  • The first (grey) goes long VXX at the close when the VIX spot will close above the front month futures contract, but the front month contract will close below the second month, i.e. VIX > VX1 < VX2.
  • The second (blue) goes long VXX at the close when both the VIX spot will close above the front month contract and the front month contract will close above the second month, i.e. VIX > VX1 > VX2.

See also trade-level statistics:

20140924.02

Evolution’s takeaway from the results is that (referring to the blue line)…

By taking into account the relationship between the VIX and the first two futures contracts, we see a significant improvement in the ability to capture upward movement in volatility.

My takeaway would be slightly different:

  • The positive results in the test of strategy #2 (blue line) are essentially the result of just two brief productive periods, coinciding with two significant volatility spikes. Out of 67 trades in total, only 28% would have been profitable.
  • It’s true that major VIX spikes will almost always eventually trigger strategy #2 because the VIX will almost always rise above the front month, and the front month above the second month in these scenarios, but…
  • These results are not sufficient reason to say the strategy is either useful or better than the alternative. Two very brief productive periods do not allow for a robust conclusion, especially considering that the limited data under consideration includes the second most significant volatility spike in VIX/VXO’s history (and so is inherently biased towards the long vol play and likely not representative of future results).

That isn’t to say that traders should never go long the VIX – we occasionally trade VXX or VXZ in our own strategy – it’s only to say that perhaps the rule presented here, VIX > VX1 > VX2 isn’t justification in and of itself to go long the VIX.

Getting the timing right on a long VIX position is extremely difficult. Whereas a short VIX position can work for a long time due to the persistent tailwind of contango (or more accurately, the volatility risk premium), going long VIX as a result of backwardation tends to only happen during major VIX spikes when the trader must combat the other force at play in VIX trading: mean-reversion. The impact of mean-reversion following a VIX spike is almost always stronger than the impact of backwardation (read more and more), hence the reason 72% of these trades failed.

* * *

A big thank you to Evolution Capital for posting their paper. This was just the first section of a more in depth discussion on their part, and I look forward to talking more about this paper in the coming week.

Good Trading,
Volatility Made Simple

Posted in Strategy Backtests.