Daily Commentary - Posted on Tuesday, November 11, 2014, 7:07 PM GMT +1

54 Comments


Nov Tuesday 11

DDN’s Volatility Risk Premium Strategy Revisited (2)

A couple of days/weeks ago I started a series of postings ( Volatility Risk Premium – Trading Volatility (Part I) ), all dealing with trading volatility ETNs / ETFs like XIV® (VelocityShares Daily Inverse VIX Short-Term ETN) and VXX (iPath® S&P 500 VIX Short-Term Futures™ ETN) and respective trading strategies. One of those strategies was DDN’s VRP Strategy (Double-Digit Numerics , Volatility Risk Premium) due to its exceptional performance – at least until December 2012 – and its compelling approach (from the paper Easy Volatility Investing from Double-Digit Numerics).

Although the strategy has delivered some extraordinary results between 2004 (data prior to the launch of ETFs/ETNs has been simulated, data source: SIX FIGURE INVESTING) and 2012, the more notable is the decline in performance since the end of 2012 (to be exact: the strategy is currently down -45.54% from its last peak in net asset value on 10/18/2012, for 629 days up to now).

This time I’ll show two different solutions to fix the recent decline in performance, replacing the VIX® (CBOE Volatility Index) by utilizing two different but related underlyings:

  • CBOE Mid-Term Volatility Index (Ticker: VXMT®), with a 6-month time horizon
  • VIX® front and second month futures (VX1|VX2), merged into a continual time series as constant-maturity futures price

But first of all the original Volatility Risk Premium (VRP) Strategy rules (always market on close):

  • Long XIV: 5-day average of [VIX index – (10-day historical volatility of SPY * 100)] > 0
  • Long VXX: 5-day average of [VIX index – (10-day historical volatility of SPY * 100)] < 0
  • Hold until a change in position.

As already discussed in some detail in my first posting of this series, using a 10-day historical volatility – additionally smoothed by a 5-day moving average – is probably too slow to react on (sudden) changes/swings in the market, while it performs (extraordinary) well during trending markets. I therefore always use a 2-day (day-to-day) historical volatility.

I wanted to follow Double-Digit Numeric’s approach not to add any additional and/or specific threshold (leave alone any additional rule) in oder to avoid “a visit from the Grim Reaper” (from the paper Easy Volatility Investing from Double-Digit Numerics).

Image I shows the respective equity curves, DDN’s VRP (original strategy) (red line), the VX1|VX2 (constant-maturity) version (blue line) and the VXMT® index version (black line), as defined by:

  • Long XIV: 5-day average of [VXMT® – (2-day historical volatility of S&P 500 * 100)] > 0
    (Please note: Before 2008, VIX® instead of VXMT® is used)
  • Long VXX: 5-day average of [VXMT® – (2-day historical volatility of S&P 500 * 100)] < 0
  • Hold until a change in position.
  • Long XIV: 5-day average of [VX1|VX2 – (2-day historical volatility of S&P 500 * 100)] > 0
  • Long VXX: 5-day average of [VX1|VX2 – (2-day historical volatility of S&P 500 * 100)] < 0
  • Hold until a change in position.

Obviously, using a 2-day historical volatility in combination with replacing the VIX® by the VXMT® not only solves the problem regarding the recent decline in performance (since 12/2012), it startet it’s outperformance already in 2008 during the financial crisis and never looked back.

Image I – Total Equity and Drawdown Curve(s)
(03/25/2004 – present)
(slippage, fees and transaction costs are assumed to total 0.1% per trade)

 

Image II shows the respective statistics for all of those strategies mentioned above.

 (click on image to enlarge)

Image II – Summary Statistics
(03/25/2004 – present)

 

Image III shows the respective monthly and annual performance (VRP with 2-day historical volatility and VXMT®).

 

(click on image to enlarge)

Image III -Monthly and Annual Returns (VRP w/ VXMT)
(03/25/2004 – present)

__________________

 

Have a profitable week,

Frank


Disclosure: I’am long/short VXX, VIX, RVX and EURO STOXX 50 volatility futures.

________________________________

Remarks: Due to their conceptual scope – and if not explicitly stated otherwise – , all models/setups/strategies do not account for slippage, fees and transaction costs, do not account for return on cash and/or interest on margin, do not use position sizing (e.g. Kelly, optimal f) – they’re always ‘all in‘ – , do not use leverage (e.g. leveraged ETFs), do not utilize any kind of abnormal market filter (e.g. during market phases with extremely elevated volatility), do not use intraday buy/sell stops (end-of-day prices only), and models/setups/strategies are not ‘adaptive‘ (do not adjust to the ongoing changes in market conditions like bull and bear markets). Index data (e.g. S&P 500 cash index) does not account for dividend and cash payments.

________________________________

Disclaimer

The information on this site is provided for statistical and informational purposes only. Nothing herein should be interpreted or regarded as personalized investment advice or to state or imply that past results are an indication of future performance. The author of this website is not a licensed financial advisor and will not accept liability for any loss or damage, including without limitation to, any loss of profit, which may arise directly or indirectly from use of or reliance on the content of this website(s). Under no circumstances does this information represent an advice or recommendation to buy, sell or hold any security.

I may or may not hold positions for myself, my family and/or clients in the securities mentioned here. Actions may have been taken before or after information is presented, and any opinions expressed in this site are subject to change without notice.

(Data courtesy of MetaStock and Pinnacle Data Corp., and for data import, testing, surveys and statistics I use MATLAB from MathWorks)

Comments (54)

 

  1. Dave says:

    excellent work – thanks

    can you explain the drivers of your model that drive your day to day change in vxx vx xiv etc? thanks again

    • TradingTheOdds says:

      Dave,

      thanks.

      What do you mean with ‘drivers’ ? It is just the formula presented, nothing else. The basic concept behind is to sell volatility when implied volatility (market expectation) exceeds realized volatility, and vice versa.

      Best,
      Frank

      • Dave says:

        Frank – which formula are you using for your daily movement. My calculations for trailing 5 day would suggest XIV and not vxx long is all I was asking. Thanks again for your help.

        • TradingTheOdds says:

          Dave,

          I thing you’re referring to the ‘optimized’ strategy which is behind the daily movement on the PORTFOLIO & TRACK RECORD page.

          I am sorry, but I’ll never discuss in detail what I am trading for my own account (having found a real competitive edge). This is a zero sum game we play. The blog is all about sharing (and discussing) ideas and concepts in order to allow my readers (and myself) to find their own niche and/or unique edge in the markets.

          Best,
          Frank

          • Dave says:

            No problem frank. Appreciate what you do. I was confused about your own proprietary stuff vs the analysis of other strategies that’s you’ve done. My model has said to be long vxx the last couple days as well and is driven by level of vix vxv spread being below the normal range ie realized vs actual vol. this is really helpful site so thanks again.

  2. James says:

    Hi Frank,
    Thanks for your work on this. Where are you getting your VXMT data. On the CBOE website, they only have historical data from 2008? Thanks.

  3. Greg says:

    Frank – I made a picture/chart based on a similar approache:
    ( http://www.gregorybarron.com/vrp )

    The top half is SPY.
    The bottom half gives a visual of how realized volatility (in lavender and using a 3-day EMA) interacts with historic volatility (yellow and orange waves), “pushing up” the VIX (pink) and longer horizon VXV (blue).

    Question – can we expect these correlations to disappear, like other market “inefficiencies”, once they are traded away? Or, alternatively, are these correlations a ‘built in” by-product of how myopic investors value risk over time.?

    Thanks for all the work on this by yourself and the other VIX bloggers you reference (ex. Vol.made.simple)!

    -Greg

  4. Greg says:

    To Dave- One way of thinking about the “drivers” is as “waves” of delayed historic volatility (depicted as the yellow lines in the chart) that form the comparative context for todays realized volatility (lavender area at the base).

  5. Marc says:

    Great work!!! Excellent.

    Have you considered trying with VXV (3-month term) rather than VXMT (6-month) or VX1&2 (1-month).

    Do you think performance would fall somewhere in between? The two charts you plotted?

    I wonder if the difference between the two equity curves is accidental driven by a few extra or fewer trades?

    Is the fact that VXMT moves slower than VIX the reason? Can the same signals be accomplished by softening the VIX instead my using a moving average on VIX or multiply it by a constant?

    • TradingTheOdds says:

      Marc,

      thanks.

      I tried with VXV (3-month term) as well, but performance was less than stellar (than with VXMT (6-month)). Unfortunately I don’t own a christal ball, but it is always possible that in the future the VXV, VX futures (either pure quotes or merged into constant maturity) or a related underlying may outperform all others due to a change in market regime. Who knows …

      I tried a lot of combinations (underlyings, moving averages, cut offs), and of course it is possible to come to even better results (than by suing the VXMT rather than the VIX) by curve-fitting (e.g. by adding a cut-off different than “0”). Currently I’am investigating into this topic (the reason behind the VXMT outperformance). Stay tuned …

      Best,
      Frank

  6. Marco says:

    It would be really interesting to understand reasons for midterm VIX overperformance. Have you tried to check for a VIX absolute value ranking in order to see if VIX sometimes beat its midterm version? Great improvement, Frank!
    http://nightlypatterns.wordpress.com

  7. kapil says:

    Hi Frank,

    I just want to make sure I am doing this correctly. For the VXMT system, the only trades that occurred in October were a sell on 10/8 and a buy on 10/15. Did you have the same trades?

    Kapil

  8. Mike says:

    Kapil,

    That’s what I got too. Long XIV from 04/15/2014 to 10/08/2014 (market on close). Then Long VXX from 10/08/2014 (market on close) to 10/15/2014 (market on close). Then Long XIV from 10/15/2014 (market on close) to present day.

    • TradingTheOdds says:

      Mike, Kapil,

      confirmed. I am afraid I misread Kapil’s statement regarding “the only trades that occurred in October …” (I thought he would be out of the market all other days).

      Best,
      Frank

  9. liam says:

    Backtested this on Canadian ETF versions. This is one case where Canada did significantly better than the American version. Still investigating why that would be.

  10. John says:

    Liam, what do you mean by “Canadian ETF versions”, and how much better did the system perform?

  11. John says:

    Frank,

    Really great work! However, I’m confused by “Image III– Monthly and Annual Returns”. The “totals” column makes no sense to me. No matter how I try to interpret what it’s calculating, the numbers don’t add up from the monthly numbers. Can you please explain? (I am treating your commas as we use periods in the U.S– to distinguish between whole numbers and fractions, of the percentage.)

    • TradingTheOdds says:

      John,

      thanks.

      It is a German Excel version, and by switching the language set there would be a few drawbacks.

      The ‘total’ column do not ‘add up’ (in a mathematical sense), monthly perfromances are multiplied (1+Jan’s performance) * (1+Feb’s performance) * … (1+Dec’s perfermance). A 50% gain in January and another 50% gain in February means you’re up +125% at the end of February ( (1+50%)*(1+50%) – 100% initial stake ), assumed you’re always all in.

      Best,
      Frank

  12. Carlos says:

    Hi Frank — why do you think XIV closed down today, Nov 14? The VIX closed down 3.48%, so shouldn’t XIV, an inverse fund, have gone UP? Instead it closed down also, by -.24%. By now XIV should be mostly composed of second month out futures, which I believe are January 2015, and they dropped today, as you would expect. So if the folks that run XIV are short these futures, I would have thought XIV would have gone up. I’m confused! Any ideas?

    • TradingTheOdds says:

      Carlos,

      XIV and VXX (as all other Volatiltiy ETF/ETNs) are based on CBOE Vix front (VX1) and second month (VX2) futures. VX1 and VX2 closed lower today, but: VX1 and VX2 went up on Thursday’s session between 4:00 and 4:15 EST (the quarter of an hour after the regular close), while VXX and XIV closing prices are based on the regular close. So VX1 and VX2 started with a backlog (approximately 0.30 index points) into today’s session. They could almost make up leeway, but couldn’t quite make it.

      Best,
      Frank

  13. Alex says:

    Frank,

    I think the new VIX calculation methodology effective since October 6th, 2014 will affect those strategies’ performance because the old and the new VIX differ sometimes as 5%. Do you think it’s better to use the old VIX?

    • TradingTheOdds says:

      Alex,

      frankly, I’ve no idea. As long as both methods differ equally on both sides – positive and negative -, any difference may be balanced out in the long(er) run. But using the VXMT instead seems to be a interesting and promising approach, and I’ll dig a bit deeper into the topic in one of my next postings …

      Best,
      Frank

  14. Jose Fraga says:

    These are the numbers I get with “VRP w/ VXMT” for 2014; performance of XIV and VXX after ETFreplay.com. I dont know why, but these results are even better as the ones exposed in Image III (98% up):
    100000
    2-31ene XIV -16,78 83220
    03feb VXX 7,09 89120
    04feb XIV 1,43 90395
    5-6feb VXX -6,41 84600
    7feb-14abr XIV 3,87 87874
    15abr VXX -1,04 86961
    16abr-8oct XIV 31,03 113944
    9oct-15occt VXX 33,37 151968
    16oct-14nov XIV 30,57 198424

    Thanks Frank!
    Jose

  15. bfft says:

    what is the 2-day historical volatility of S&P 500? I dont understand this part

  16. bfft says:

    5-day average of [VX1|VX2 – (2-day historical volatility of S&P 500 * 100)] > 0

    I don’t understand what you mean by the 2-day historical volatility is. Confused. could you give an example

  17. Gregg says:

    Frank, what granularity do you use to measure the SPY’s trailing volatility? Minutes, hourly, etc?

  18. solo says:

    I found combining with a simple indicator (not tuned) would turn $10k to $3.7mil in 2.5 years in back testing the Canadian versions. quite surprising.

  19. QuantTrader says:

    Hi,

    Interesting. I am trying to backtest in Zipline (Python). Some help is appreciated. I also like to turn $10K into $3.7 mln in 2.5 year :-). lol.

    1. How can I compute a ‘VIX® front and second month futures (VX1|VX2), merged into a continual time series as constant-maturity futures price’?
    2. I read you use S&P 500 index and not SPY.
    3. Is there a way to protect against drawdown ?
    4. How to calculate the terms structure based on Vx1,Vx2? Maybe off-topic, but read in other article that this could help in predicting when to be neutral instead of in the market for contango/ backwardation.

    J.

    • TradingTheOdds says:

      Hi,

      1. simply take into account the closing values of 1st and 2nd month futures, the total days to maturity (previous month to current month, regularly between 19 and 22) and the respective remaining days to maturity. Or look here :http://www.ipathetn.com/US/16/en/details.app?instrumentId=259118 (daily values)
      2. Correct.
      3. Yes – put your money under your matress. ;-) Otherwise potential losses are always a function of risk. But if you find a way to avoid drawdowns, please let me now …
      4. I don’t make use of the term structure.

      Frank

  20. QuantTrader says:

    2. If I use ^GSPC instead of SPY from Yahoo I get a VRP between [0 … -100]. The long/short on VRP>0 does not work then. It looks there is a kind of offset. How to elimate that?

  21. Ovidiu says:

    Hi Frank,

    According the this strategy, there hasn’t been any switch since October, and the results are

    Nov 2014 -21.24%
    Dez 2014 -11.27%
    Jan 2014 – 8.71%

    What do you think about this ?
    How long can the VIX stay overbid and SPY make such small moves ?

    For the next 3 months.
    This is the first time when the strategy is 3 months in drawdown, since 2004.

  22. Fritz says:

    Hi Frank
    Thanks for reporting the results honestly.
    Regards
    Fritz

  23. alex says:

    Interesting article, thanks for sharing.
    I am wondering if using vix in life of vxmt pre-2008 isn’t a problem though : on the one hand you use short term vol, on the other hand mid term.
    Have you tried to find time the system with tailing stops and take profits to mitigate drawdowns? (I usually analyze MAE and MFE to cut losses and remove outliers)

    Had anyone found the Canadian equivalent someone was talking about by the way?

    • TradingTheOdds says:

      Alex,

      the problem with backtesting trailing stops is to have intraday data for XIV and VXZ (time and quotes) in order to check if a buy or sell stop had been trigger first (sometimes XIV and VXX move big during the day), and if the system would’ve triggered an intraday buy again after being stopped out during the sessions. I prefer to keep it simple.

      Best,
      Frank

  24. alex says:

    Hi Frank, Thanks for your response.

    Could one alternative be to use delayed entries?
    My (basic) algo uses a mix of trend (moderate or strong contango : long xiv) and countertrend signals (vix above a certain level: go long xiv). It is a long only algo (I do not like vxx). Signals are calculated on MOC but entry/exits are MOO. (and living in Asia it is pretty suitable as I am awake when market opens and sleeping when market closes).

    Flip side is that when trailing stop set at x%, with MOO, the loss can exceed its set stop…

  25. sanpal says:

    Hi Frank,

    Excellent post! I have been playing with vxx, xiv switching strategies for a while now. Just like what you showed here, I have had some success in optimizing the VRP strategy to get rid of the lull in the recent years (2013+). However, I find it almost impossible to improve returns from start of 2007 to August 2008. This has been my biggest reservation about the xiv/vxx strategy. I see your best performing strategy (VXMT) suffering from the same problem. Do you know what is so significant about this time period? Have you had any success optimizing for this period? Any ideas would be much appreciated.

    Great work!
    Thanks!

  26. sanpal says:

    Thanks to the link Frank! This is very helpful. Looking forward to reading more of your posts.

  27. Cap says:

    I noticed here and at

    http://volatilitymadesimple.com/chasing-the-volatility-risk-premium/

    the SPZ HV10 is used in 4-VRP whereas in the paper

    http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2255327

    VIX HV10 is used (or am I missing something?). Why was it changed?

    • TradingTheOdds says:

      The Volatility Risk Premium Strategy always uses the S&P 500 (or SPY) historical volatility (smoothed by an x-day SMA/EMA) in comparison to the CBOE Volatility Index (in both papers/articles). VIX HV10 represents either the (annualized) 10-day historical volatility of the VIX (regularly) or the 10-day simple moving average of the VIX, depending on the notation of the author.

  28. Len says:

    Frank,
    For clarification: Using the macroption.com calculator, I plug in data for GSPC and use the 2 day lookback period with Std Dev formula. This will give me the S&P 500 2 day historical volatility?

  29. Len says:

    Using Zero Mean 2 day historical volatility I get a cross over from XIV to VXX on August 21, 2015 and cross back on September 10, 2015. Is this correct?

  30. flying Frenchman says:

    hi,how has the strategy performed since it was posted? could you share a spreadsheet helping us with the calculations please? thanks! alex

Leave a Reply

Your email address will not be published. Required fields are marked *