Seeking Alpha

Fred Piard's  Instablog

Fred Piard
Send Message
I design systematic strategies for clients and myself. I trade stocks, sectors, countries, bonds, volatility. I publish Newsletters mirroring my own portfolios. My books explain simple, low-risk strategies: "Quantitative Investing" got 5 stars at Amazon, "The Lazy Fundamental... More
My company:
Ypa Finance
My blog:
Ypafi Global Household Index
My book:
Quantitative Investing - Strategies to exploit stock market anomalies for all investors
  • XIV Returned 41% A Year For 3 Years: How To Make It Better With Less Risk. 8 comments
    Jan 16, 2014 11:35 AM | about stocks: SVXY, VXX, XIV

    "Risk, I had learned, was a commodity in itself. Risk could be canned and sold like tomatoes."

    Michael Lewis

    Trading volatility derivatives looks more like scientific gambling than like investing. First, because considering volatility as an asset class is questionable. Second, because VIX ETNs are riskier than stocks and equity ETFs. However, they have characteristics that make them very profitable in certain market conditions.

    Knowing the risks.

    The VelocityShares Daily Inverse VIX Short-Term ETN (NASDAQ:XIV) is a risky product by design. It is very volatile, termination events are defined from inception (reading the prospectus before buying is always a good idea), and as an ETN it incurs an additional counterpart risk. Moreover, it is one of the most "derived" of derivatives. Imagine an ETP, which is inverse, derived from futures, derived from the VIX index, which is an abstract aggregate calculation on options, derived from stocks of the S&P 500 index. It means that the links with usual market laws are indirect. As a consequence, usual analysis techniques might not work exactly the same way. When it comes to building a trading model, XIV backtests are very short. Usually I backtest stock strategies on 15 years. But VIX futures started in 2004, VIX ETNs in 2009, and XIV on 11/29/2010. Since October 2011, there is an ETF equivalent to XIV: SVXY. It removes a counterpart risk, adds the possibility to trade options (the derivation level gives me headaches), but has a shorter history and a lower trading volume.

    Understanding the edge.

    So, why bother trading a risky and atypical product? Because it has a very special bias: it allows to win even when being slightly wrong. This is because of the term structure of underlying VIX futures, which is usually in contango. The concept is easy to understand for commodities. Contango appears when buyers are ready to pay a premium for a later delivery. Just like a factory director anticipating higher raw material prices would try to buy stuff now, and rent a warehouse space to store it. Contango can be assimilated to the warehousing cost. The opposite situation, called backwardation, can be interpreted as a premium for fast delivery. I agree it is difficult to imagine volatility being stored and delivered. That is just an analogy. But the fact is, VIX is most of the time in contango. It explains why, just by rolling short positions in future contracts, XIV goes up even when volatility remains in a range. About ten millions XIV shares are traded per day. There is no problem of liquidity, spread and slippage for an individual investor.

    Designing a strategy.

    I limit the risk by playing XIV only on the long side (its bias direction), without using margin, and with a reasonable amount of money. As I have no vocation to become a day-trader, I am looking for weekly setups. The game is to take a decision every Monday: buying XIV when the VIX index is likely to decrease or stay stable, selling or staying in cash when it is likely to rise. I like keeping things as close as possible to common sense. It doesn't mean that all my models are simple: they may be sophisticated logical combinations of simple rules. In this case, I tried to combine technical analysis and inter-market behavior. After a few months of research, observation and real trading, I was able to produce a model showing this following 3-year simulation chart (1/1/2011-1/10/2014):

    (click to enlarge)

    This is a rule-based model. The rules, mainly technical, cannot be disclosed here. The next table compares it with buying and holding XIV all the time (B&H).

     

    Annualized Return

    MDD*

    Sharpe Ratio

    STD**

    Correlation / S&P500

    XIV B&H

    41.43%

    -74.55%

    0.49

    80.44%

    0.50

    XIV Model

    77.00%

    -10.88%

    2.46

    30.39%

    0.11

    *Maximum drawdown - **Standard Deviation (risk)

    A 0.2% rate is accounted for trading costs. The performance looks great, but 3 years is a short time for a backtest, and the simulation covers only 20 trades (40 signals).

    Testing the model.

    In a next step, I reversed the rules so as to build an equivalent short-selling model with VXX. XIV is almost the inverse of VXX, and the latter can be backtested on 20 additional months. The aim was to perform an out-of-sample test on these 20 months. Performing an out-of-sample test before the design sample period is quite uncommon, but it is an interesting possibility.

    Here is the VXX reversed-model simulation results for the out-of-sample period (4/1/2009 - 1/1/2011), and a comparison with selling and holding VXX short (S&H):

     

    Annualized Return

    MDD

    Sharpe Ratio

    STD

    Correlation / S&P500

    VXX S&H

    68.74%

    -94.14%

    0.29

    223.12%

    0.26

    VXX Model

    38.12%

    -18.02%

    1.07

    32.55%

    0.31

    This period represents 15 additional trades (30 signals). Performance is less impressive, but still good. The annualized return is lower than holding VXX short, but it avoids the excessive drawdown and volatility.

    Here is the equity curve of the VXX model for the whole available period (4/1/2009 - 1/10/2014), with an annualized return of 47%:

    (click to enlarge)

    My next step was to compare XIV and VXX models on their overlapping period (1/1/2011-1/10/2014). The results are in the next table:

     

    Annualized Return

    MDD

    Sharpe Ratio

    STD

    Correlation / S&P500

    XIV B&H

    41.43%

    -74.55%

    0.49

    80.44%

    0.50

    VXX S&H

    54.65%

    -77.33%

    0.63

    83.50%

    0.50

    XIV Model

    77.00%

    -10.88%

    2.46

    30.39%

    0.11

    VXX Model

    52.77%

    -12.85%

    2.02

    25.02%

    0.11

    The "hold" strategy is better with VXX, however the model works better with XIV, which is good news for people who are afraid of short-selling (with good reasons).

    On the overlap, the XIV model has an overhead of 24% on the VXX model. If we translate this overhead to the out-of-sample period, we might guess a return about 62% if XIV had been available. I don't take this number for granted: it is supported by a generalization, not by a scientific rationale. It is just a clue that the XIV model may have done better than the VXX model in the out-of-sample test, if XIV had been available at this time.

    Conclusion.

    A simulation since April 2009 is very short. It includes a flash crash (May 2010) and a 20% market correction (August 2011), but no bear market. Despite this, and despite being a conservative investor, I consider that this model looks good enough to bet less than 5% of a global portfolio with a close monitoring. The historical maximum drawdown is below 20%, which gives an alert level to put it in quarantine with a global loss under 1% if it crosses the limit.

    In theory, it would be possible to translate the model in future contracts and to simulate it since 2004. The VIX index has been calculated before 2004, however a backtest before future contracts data would need a deterministic model of contango and backwardation. As they are driven by psychology and not a calculation, I doubt it is possible. And, even if we had decades of data, past performance, real or simulated, is never a guarantee for the future.

    Disclosure: I am long XIV, . I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

    Additional disclosure: I might have already sold my long position in XIV when you read this.

Back To Fred Piard's Instablog HomePage »

Instablogs are blogs which are instantly set up and networked within the Seeking Alpha community. Instablog posts are not selected, edited or screened by Seeking Alpha editors, in contrast to contributors' articles.

Comments (8)
Track new comments
  • jacobtr
    , contributor
    Comments (291) | Send Message
     
    Nice article, and good job tweaking an already impressive idea.
    16 Jan, 12:41 PM Reply Like
  • AmSO4Dave
    , contributor
    Comment (1) | Send Message
     
    Thanks for sharing your thoughts. My concern is what happens in a Bear market? What conditions would create backwardation?
    17 Jan, 02:32 PM Reply Like
  • Fred Piard
    , contributor
    Comments (589) | Send Message
     
    Author’s reply » Backwardation usually happens when the VIX spikes. It is quite rare, and on short periods. Some of the rules are designed to go out when volatility goes up beyond an acceptable pace. I don't really fear a bear market, I think the main risk is a flash crash. There was one in May 2010 and the reversed VXX model did well. A single event is not a proof, but a flash crash has a higher probability to happen when the market is already nervous, and the model out of the market. All is about probabilities, nothing is sure.
    17 Jan, 04:28 PM Reply Like
  • MarcJoli
    , contributor
    Comments (78) | Send Message
     
    Fred, good articles. I am also long XIV most of the time. My trading rules include whether the term structure is favorable to XIV. I have found some basic chart indicators work very well for timing entries and exits. I have one strategy in my instablog and planning to write an update soon.

     

    @ContangoMojo
    13 Feb, 01:59 PM Reply Like
  • Fred Piard
    , contributor
    Comments (589) | Send Message
     
    Author’s reply » Hi Marc, I will have a closer look. My signals are weekly, and the model described above has the opposite philosophy: being out of the market most of the time. Both may work.
    13 Feb, 08:58 PM Reply Like
  • User 12669251
    , contributor
    Comment (1) | Send Message
     
    Where may your "rules, mainly technical" be disclosed?
    13 Feb, 03:48 PM Reply Like
  • Fred Piard
    , contributor
    Comments (589) | Send Message
     
    Author’s reply » They are private. Some subscribers follow my signals.
    14 Feb, 06:31 AM Reply Like
  • Macro Economist
    , contributor
    Comments (377) | Send Message
     
    Fred, I have the underlying benchmarks on VXX and VXZ since December 2005 if you want to PM me. I can backtest your strategy for you (I have my own).
    28 Apr, 02:05 PM Reply Like
Full index of posts »
Latest Followers

StockTalks

More »

Latest Comments


Posts by Themes
Instablogs are Seeking Alpha's free blogging platform customized for finance, with instant set up and exposure to millions of readers interested in the financial markets. Publish your own instablog in minutes.