OptionVue's VXX Trading System

From the article I referenced above:-

"There is a complication in the futures market that we must talk about because many people confuse it with the risk premium. It is the roll yield. There is a widespread belief that profits from the VIX futures markets come from the roll yield. This is not true because
  1. (i) if it were true then the roll yield would be a free lunch
  2. (ii) the profits actually come from the risk premium

    This is explained in detail using numerical examples by Gorton & Rouwenhorst (2004). The roll yield is the difference between the spot VIX and the futures price. It is called a yield because it may pay or cost a small amount every day as the futures price and the spot VIX converge on each other.
    For example if the spot VIX is $14 and the futures price is $18 and an investor has a short contract then every day as settlement approaches the price of the contract drops a little – e.g. $17.80, $17.65, ... so the short investor gains a few cents each day.
    But remember that the roll yield, per se, is not a free lunch. The reason that the futures price is $18 is because the VIX is predicted to rise to that price. So the price may stay at $18 while the VIX rises to meet it – e.g. $14.26, $14.37, ... In that case even though the roll yield still exists it does not pay off because the futures price stays at $18 without moving. We emphasize this point that it is the VRP, not the roll yield, that provides the profit. If there were no VRP then there would be no profit"

Posted by DavidF
 
David, Ryan,

I understand the perspective that you and your referenced papers are suggesting, but to broadly state that "the returns don't come from the roll yield but instead come from the volatility risk premium" fails to acknowledge that any volatility risk premium - were it to exist (and the existing data suggests that it does exist) - is imbedded in the roll yield.

Introducing a little bit of math to illustrate the point (the example here only considers short VX trades when the term structure is in contango):

VX(t,T) = VIX futures at time t, expiring at time T
VIX(t) = VIX at time t
VIX(T) = VIX at time T > t
E[VIX(t,T)] = expected value of VIX(T) at time t

The P&L of a short VX position at time T, which was entered into at time t is:

VX(t,T) - VIX(T)

This can be disaggregated as:

VX(t,T) - VIX(T) = [VX(t,T) - VIX(t)] - [VIX(T) - VIX(t)]

The first term, [VX(t,T) - VIX(t)], is the roll yield. The other term, [VIX(T) - VIX(t)], is the change in spot VIX. When the term structure is in contango, the roll yield is positive (once again, assuming your position in the futures is Short). If, as the historical data suggests, E[VX(t,T) - VIX(T)] > 0, then it seems obvious to me that the positive roll yield contributes meaningfully to any positive expectancy that may exist by having a short VX futures position.

The idea you have put forth is to disaggregate the roll yield into a VRP component and an expected drift component, to try to answer the question "WHY does this positive roll yield exist?":

VX(t,T) - VIX(T) = [VX(t,T) - E[VIX(t,T)]] + [E[VIX(t,T)] - VIX(t)] - [VIX(T) - VIX(t)] = [VX(t,T) - E[VIX(t,T)]] + [E[VIX(t,T)] - VIX(T)].

[E[VIX(t,T)] - VIX(T)] is noise with expected value 0, so the expected return can be defined as [VX(t,T) - E[VIX(t,T)]], or the volatility risk premium. I get it.

But really guys, you're splitting hairs for a purpose that I cannot understand. The answer to the question "Why did my 401k balance go up today?" could be (1) the individual mutual funds I own went up in value today, or (2) on average, the individual stocks and bonds that my mutual funds own increased in value today. Both are correct, and it seems silly to spend any energy arguing that " (1) is incorrect and (2) is correct, because (1) does not happen without (2) necessarily happening".
 
Most of this stuff is way over my head.
I played around with trying to merely visualize the difference in the front and back month VX futures, to get a sense of how the contango would impact SVXY or XIV.
Here is a TOS Thinkscript that merely examines the difference, if you are curious.
.txt appended to the name to allow posting here. -- it is a thinkscript text file.
 

Attachments

KJ, yes, if you made money you made money, doesn´t matter why. But I think it´s important to understand what you´re actually betting on when you put money into XIV.

If you understand how it makes money you can then ask yourself questions such as "are the VX futures underestimating a potential spike in spot VIX approaching a binary event" or "are VX future expiry dates in relation to upcoming events influencing the curve". Then you´re making a more educated decision as to what you´re betting on, i.e., that the futures market is overestimating where spot VIX will settle.

Posted by DavidF
 
If anyone´s interested here´s a comparison of XIV strategies from 2004 to 2013. The VRP strategy is optimal, as the authors point out the weakness of basing a strategy on "roll yield" is that VIX is mean reverting not only from high but also low levels. Roll-yield still works well compared to buy and hold, but that´s due to a high VRP being associated with a greater roll yield. The "roll yield" strategy is if the 10 day moving average of VXV/VIX > 1 go long XIV else go long VXX, and the VRP strategy is i"f the 5 day moving average of (VIX - 10 day historical volatility) > 0 go long XIV else go long VXX".

I think OVs strategy should beat the roll-yield strategy above (not good to always be in either XIV or VXX) and in combo with the Yates Danger Signal it might also incorporate aspects related to volatility of VIX that are covered in the VRP strategy.

Screen Shot 2016-12-09 at 13.00.10.png

Posted by DavidF
 
This website used to follow 24 different volatility strategies:
http://volatilitymadesimple.com/blog/

Does anyone have an access to Trade Station or other back-testing software and could run a quick automated back-test for this very simple strategy for period from 2010-2016 ?
1) Go long Volatility ( buy VXX) at the close the day when VIX is about to close above VXV on hourly chart.
2) Go short Volatility ( sell short VXX or buy XIV) at the close when VIX is about to close below VXV on hourly chart
 
I am not a volatility trader, so please take these comments with a grain of salt. Len's VXX system has been successful, but here's food for thought. I recently spoke with Karen at OV about another matter. Since we are "old friends", we briefly discussed other matters including the VXX system. I understood her to say that his system only made four trades this year, as one might expect. Also, she said that the draw downs can be shocking to some people, even though "it always has come back" (her paraphrased words). I always try to interpret everything in the day-to-day and here-and-now (possibly a consequence of being a physician), so this strategy does not suit my trading style. But, I am very conservative (and have the gray hair to prove it) and really hate draw downs of any magnitude. Having said all this, I think his track record so far has been good. Apparently, he also trades OPM with this system.
 
I second what Dan says about the drawdowns.
If you ever decide to trade volatility, you have to be open-minded and be ready to experience a 50 % drawdown in 3-5 days.
Therefore, the only solution to sleep well at night is to trade very small and allocate max 1 % of your net worth to any volatility strategy.
Before you put any large capital at risk, I suggest at least to look at XIV/SVXY daily chart for August 2011 or August 2015.
I made very good money shorting volatility in 2012-2014 but completely stopped trading volatility after last August.
I didn’t blow up my account only because I was hedged with 10 % OTM weekly SVXY puts but still lost a lot of money.
Kevlar type or M3 style trades are much more sleep and heart friendly and I am a much happier person now than at the time I was trading Vol.
Just my 2 cents.
 
I finally had a chance to read the paper that has been referenced multiple times in this thread. For anyone who has used this paper as a basis for any of their trading strategies/decisions, I would caution against taking the author's conclusions as gospel. The author appears quite comfortable in waving his hand over some of the important math in reaching his conclusions.

As an example: The author suggests that realized VRPO has some predictive power to forecast future VRPF because the historical correlation between the two is sufficiently high (in the author's estimation). This reasoning is quite weak and unconvincing to me. The battlefield of investing is littered with failed strategies that rely on tenuous correlations as a means to forecast the future. As an exercise in bringing the analysis forward, I would recommend that you see how effective the HVOL10S decision rule has worked since 2013. It looks to me that the predictive power has completely disappeared.
 
If Dan´s words are to be taken with a pinch of salt then mine can be taken with a kilo of salt given my experience vs. his, but I don´t see a clear distinction between "volatility traders" and what we do, we´re all selling volatility risk premium on indexes. Only difference is we have elaborate software to help us control risk with delta and provide a reasonable predictive visual.

With a beta of around 3.8 or so, you have to understand that if you put your 100k account into the VXX system on a buy signal, it´s not too different from putting $400k in the SPX, and a 10% SPX drawdown could wipe out most of your account (or a spike in VIX without a big fall in SPX). And not too different from trading an M3, Kevlar or RTT that´s far too big for your account.

AKJ, not sure how you interpreted what you did, he wrote that the correlation between VPRO and VRPF was 0.56, and that you can´t expect it to be higher as there´s no reason to believe the 2 should be correlated. But that there is a correlation (granted no stats mentined) and hence one provides some info about the other.

XIV obivously gains most when VRPF is high and VRPO is low, unclear to me why people bid up VX futures and not SPX options but in periods like this summer it´s striking.

For the record I don´t agree with his defintion of "roll yield" either but it´s useful for studying the ability of VX2-VX1 to predict XIV gains. What I don´t agree with (and the point of my original post) is the "and the fund owners are forced to buy more expensive further dated futures", as there´s obivously no loss of value in the roll, you just get less futures for your money. I it was just loss fof value with time option decay could be called "roll-yield"

http://lexicon.ft.com/Term?term=roll-yield

Posted by DavidF
 
tom said:
AKJ, not sure how you interpreted what you did, he wrote that the correlation between VPRO and VRPF was 0.56, and that you can´t expect it to be higher as there´s no reason to believe the 2 should be correlated. But that there is a correlation (granted no stats mentined) and hence one provides some info about the other.
Click to expand...


...sigh....you are making my point for me, but do not recognize it. Empirically observed covariation is a necessary but not sufficient condition for causality. Your statement that "one provides some info about the other" MAY be correct, but has not been sufficiently shown in the paper. Simply stating that there is a 0.56 correlation is insufficient in showing that realized VPRO can be used to predict future VRPF.
 
Incorrect, AKJ it wasn´t my statement, HE wrote that one provided info about the other. I therefore assume that it was statistically validated. Let´s drop this for everyone´s benefit.

Posted by DavidF
 
So I´ve been thinking about what you´re actually buying/selling with VXX/VIX versus futures where there´s a physical underlying, e.g., crude, and I concede that my claim that changes in value are not associated with "roll yield" is largely semantics.

My original thinking was this:- you want exposure to crude oil so you buy a front month CL contract giving you 1000 barrels for $50k. The term structure has a approx 5-10% increase in price evey month, and the spot converges up to the next future every month. To maintain your 1000 barrels, the roll costs $5k++ a month. In 10 months oil is $100 but you´ve actually lost a lot of cash due to negative roll yield.

If instead you wanted to be long volatility and you put $50,000 into VXX, and the term structure was identical to CL, i.e., a 5% increase every month with spot converging up to the next month, after 10 months you´d still have $50,000 as you don´t plough more capital in to maintain the same exposure. Hence, ostensbily at least, the absolute value of VXX/XIV price is not driven by positive/negative roll-yield (my former reasoning).

However, even if you haven´t lost capital. to argue that "roll-yield" doesn´t influence value of VXX/XIV is the equivalent of saying owning 500 barrels of oil for $50,000 is the equivalent to owning 1000 barrels for $50,000 10 months prior.. It´s an illusion and the difference is only related whether you want to manintain the same no. of futures contracts (trading VX futures) or only use existing capital (VXX/XIV ETFs) at each roll. You lose value either way. So I stand corrected.

Posted by DavidF
 
Thanks SVL, good article and pretty much covers disagreement on definition, also in comments sections.

I´m still not sure what controls VX futures price vs corresponding SPX IV along the chain. With low IV and steep contango, speculators are bidding up the price of VX futures in relation to spot VIX.However there doesn´t seem to be the same risk premium in the further dated SPX options.

Posted by DavidF
 
Lot of interesting discussion in this thread, both theoretically and practically. The geek in me worked out some math too, but I won’t bore you here.

Practically, I coded the 2 strategies published in Cooper’s article last year (along with Connors-Avarez VXX strategy, but it didn’t work well so I discarded it). VRP suffered a big loss last August. Roll Yield strat has performed better. Currently I’m using both, not verbatim, but as a timing/position sizing scheme for my VXX BWB trading.

Attached are the results for VRP on VXX (long and short), from 2010 to 2015 and 2015 to 2016 separately.
I will compare 2 strats side by side from 2010 to 2016 and post the results here

Posted by Nam
 

Attachments

Thanks Nam, good info. Are the VXX BWB not very commission intensive? Would be great to hear more.

For myself the optimal way to manage risk-reward is combos with short SPX/long XIV set-ups. The formulas based on VRPO/roll yield moving averages etc are, IMHO, blind to the complexities of volatilty pricing and term structure. It´s the equivalent of selling puts on a stock based on an IV differential value that doesn´t take into account factors such as IV ranking, earnings or historical moves in relation to IV.

The 3.8 beta of XIV gives you a ballpark ratio of the correct ratio and a 1:4 ratio of short ES to XIV performs pretty well to both the downside (profit Aug 2015) and upside if you can select the correct periods , i.e., when it´s good to go long XIV it´s normally good to use this ratio. Granted you give up gains to the upside in XIV but if you can catch periods where the term structure is steep, SPX trades flat and you rebalance by taking profits, it´s a great set-up. Hits are spikes in IV with no drop in SPX but that shouldn´t last, either SPX drops or VIX reverts.

Posted by DavidF
 
Attached is the comparison for the VRP and RY strategies. Codes are also included. They’re in Amibroker, and simple enough to replicate in other languages. I tested the strats using EOD data. I also tested couple of other systems (from well known vendors) but these 2 perform the best

DavidF: if you trade VXX and VIX BWB as an “income” trade, then commission is a headache. I use indicators to establish bias and make max 2 adjustments during the life of a trade so transaction cost is bearable. The indicators I use are (in order of importance)
1-VIX spot and its 10 day MA, Bollinger Band
2-VIX/VXV (i.e. Roll Yield strat)
3- IV-HV where HV means 1M historical vol, 10 D HV (VRP strat), exponentially weighted HV, HV based on H L

When VIX approaches its upper BB, I start to scale in, usually 1/3 of capital. I use call BWB and adjust the delta depending on my perception of the risks (some human judgment is involved). I used to try hard to keep theta positive but now I don’t care anymore. Theta can be negative. VIX and VXX Flys hold their values well up until 2 weeks before expiry. Then I watch VIX/VXV. When it gives a signal to go short I enter another tranche or move the short strike of the first tranche closer to VXX, depending on the strength of the signal. After the 2nd tranche, if the 3-type indicators give strong entry signals then I enter a third tranche. This system involves some discretion but mostly systematic. As compared to naked short VXX, BWB has lower drawdown, but still large. So now I scale in. sometimes I miss opportunities but that’s fine by me.

Posted by Nam
 

Attachments

Thank you for sharing Nam, excellent overview. If I´ve interpreted the VRP and RY data correctly over the time period tested it never paid off to be long vol. using the rules for either strategy. As I mentioned above the RY strategy is similar to OV´s old VXX trading system when you´re never in cash.

One question regarding your BWBs, did you formerly trade 2-legged debit/credit spreads but found BWBs gave better returns?

Posted by DavidF
 
You must log in or register to reply here.
Top
Contact Us