Strategy #2 for Trading Volatility ETPs: Timing the VIX
This week I’ve been talking about 2 strategies for trading volatility ETPs like VXX and XIV (read more and more). In this post I’ll drill down on the second of those strategies: timing the VIX.
As I’ve shown before, timing the VIX index itself is easy (albeit useless):
In the graph above I’ve assumed we could buy/short the VIX index at the close when the 10-day EMA of the VIX closed under/over the 10-day SMA. An EMA is a faster moving average than an SMA, so this is a mean-reversion strategy.
The graph shows that the VIX is uber predictable.
The problem of course is that back here in the real world the VIX futures term-structure acts as such a strong head/tailwind on volatility ETPs that it effectively becomes the most important driver of returns, not movements in the VIX itself.
But what if we traded these ETPs by timing the VIX, but only taking those trades where the term-structure put the trade in our favor?
The Strategy
Go long VXX/XIV at the close when the 10-day EMA of the VIX closed under/over the 10-day SMA, AND the average daily impact of the term-structure was greater/less than zero, otherwise move to cash.
See this post for more on estimating the average daily impact of the term-structure. Strategy results from 06/2004 (frictionless):
[logarithmically-scaled, growth of $10,000]
Despite its simplicity, this strategy would have put up impressive numbers: a 53% annualized return with less than 1 trade per week and a max drawdown of -33%.
Like the strategy I presented in my last post, there’s clearly some low hanging fruit to improve on these results (like stop losses), but the smoothness of the equity curve through very different volatility regimes demonstrates nicely that timing the VIX while respecting the VIX futures term-structure is an effective strategy.
As I talked about previously, I use a mix of both types of strategies.
Following the term-structure (ex. strategy #1) ensures I always have some exposure to which way the winds of the VIX futures term-structure are blowing. And timing the VIX (strategy #2) tells me when the market is particularly ripe for a trade so that I can up my risk exposure.
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Trading Strategies, VIX & Volatility | 16 Comments
Great posts, Michael.
Great to see you are back publishing.
Even better than yesterday!! Really, REALLY interesting.
Great work !.
Question: where did you get the historical data for the term structure ?
Thanks
eber
Hello Eber – this post from the Intelligent Investor Blog can help with historical data for the first and second month VIX futures:
http://investing.kuchita.com/2011/08/21/vxx-data-since-vix-futures-avilable-march-2003/
Note that for my estimate of (what I loosely call) the “term-structure”, I’m extrapoloating from the actual ETP and VIX prices. But using the VIX futures themselves (like the blogger does in the link above) is going to give you a very similar result.
michael
Although the results look impressive I get the feeling the environment over the past couple of years has been very hospitable to this trade and it would be dangerous to assume past will match future.
The returns from 2004-2008 are only about 16% annualized, with the bulk of the gains coming during the huge bull market beginning in 2009. Since 2009 the market has been tightly coupled to central bank liquidity. It goes up remarkably smooth when central bank liquidity programs are running and is highly mean-reverting when negative shocks are encountered when QE programs are running. If you’re trading the VIX EMA/SMA mean reversion in a bull market you are basically betting on QE to cause mean reversion after a negative shock happens. I get the feeling this environment is not sustainable and when the liquidity games stop this trade will become much less profitable and much more dangerous both because of the increased return volatility and not being able to detect the regime change until much later.
I think if you look at your data and instead buy SPY at the times you would have bought XIV you’d get significant excess returns there as well, although not as much because VIX has natural ~3x leverage built in.
RE to Jeff: couple of thoughts:
First, the mean-reversion in the VIX is clearly not tied to QE. See the first chart in this post. Note how the strong tendency towards MR in the VIX has existed since its first quote in 1990.
The improved performance 2009+ has nothing to do w/ an increase in MR, it has everything to do with consistently contangoed VIX futures, which to your point, one could argue is a result of QE.
All straight long/short volatility ETP plays are going to be less productive when the direction of the term-structure is unclear, and without a doubt, I don’t think anyone could expect big returns all the time.
Having said that, annualized returns for this simple strategy pre 2009 are in the neighborhood of 24% (I’m not sure where you’re getting 16%). Couple that with the fact that this is an overly simplified strategy without the slightest bit of sophistication, and I think it’s pretty safe to say it has wings (at least in my humble opinion).
P.S. I disagree re: trading SPY. That would be a straight MR play and usually is less productive. For better or worse, the primary driver of the strategy I’ve detailed in this post isn’t MR, it’s the term-structure.
michael
How much of the returns are due to leverage, would you say?
Thank you Michael.
RE to JZ: there’s no “leverage”. Jeff means that these products are more volatile than the S&P 500.
That’s why I always say that “returns are an illusion” – they’re just a function of exposure/leverage/volatility. It’s much more important to look at the smoothness of the equity curve and volatility-adjusted measures (Sharpe Ratio, Ulcer Index, etc.)
michael
Very nice strategy, I’ve developed something very similar.
Do you have more statistics to share? volatilty, sharpe, sortino ratios?
have you tried to assume trading and borrowing costs to see what woudl be the impact on performance?
Thanks Eugenio
Hello Eugenio – I don’t trade this strategy, it’s just an illustration, but I trade something conceptually similar. I didn’t bother with my usual statistical breakdown b/c I would never suggest anyone trade the strategy as is (especially without stop losses), but if you did: annualized return = 53%, annualized SD = 37%, Sharpe Ratio = 1.4, max DD = -33%, winning months = 68%, days in market = 50%
Trading costs would be negligible assuming a decent sized account (less than 1 trade per week). There are no borrowing costs because these are ETPs sans margin.
michael
Michael, thanks for sharing your great work on this subject!
The volatility ETPs indeed open up some potentially lucrative strategies. I’ve developed a similar system that has worked beautifully so far. Here’s hoping the good times last.
best,
Mike
Dear Michael,
I have a question about your VXX/XIV strategy progamme. As you know VIX is experiencing from time to time intermediate expansions (please see here for example: http://av.r.ftdata.co.uk/files/2012/08/Vix.jpg). Based on history (that’s what we all do in quant analytics) I am expecting volatility to expand in the intermediate term any time soon. Would you be so kind and provide me with a split of the performance of your strategy in times of rising volatility (like August 08- Oct 08, April 10-July 10, July 11- Oct 11) and contracting volatility? Thank you
Hello Krystian – I never release backtested results for any of our proprietary strategies. Goes against everything I believe folks should be using to judge the efficacy of a strategy.
Having said that, great question. I’m not sure by your sentence if you’re saying that you do or do not expect vol to expand.
In either case, the biggest risk to the strategy is a sudden, major change in volatility (especially a spike upwards). We can do things to try to control that risk like using protective stops, monitoring futures term-structure, and timing the VIX, but at the end of the day it’s an inherent unknown that all directional volatility plays face.
My intent is for the strategy to underperform a strong bull market (and consistently contangoed futures) simply because any trade other than fully long XIV is going to underperform. But my intent is also that we’re able to minimize losses and continue to generate returns when the VIX spikes and the term-structure turns.
That spike hasn’t happened since I began officially tracking the strategy so I understand it’s an unknown for readers. Unfortunately, that’s not sufficient reason to break my “no backtest” rule (sorry).
michael
Dear Michael,
Thank you for posting this and the other fantastic strategy (i.e. #1)! Just wanted to let you know that I have been able to replicate and improve on the basic models you shared using stops and a few other simple tweaks. I have been able to backtest results to 2004 using simulated VXX and XIV prices and have confirmed the more recent simulated history to be valid using acutal VXX and XIV prices. The raw models I put together are a bit scary in terms of drawdowns but with protective stops they become much more reasonable and the stop values can be over a wide range and still produce a level of risk that I can sleep with.
Looking forward to a short period of paper-practice trading to ensure I can execute the orders on time, then getting started with some real money!
Absolutely great stuff – I can’t thank you enough for sharing.
Wes