VIX and SPY Show Positive Correlation For 4th Day In A Row

VIX Futures down slightly today, but remain largely unchanged at the close for 4 days now. The S&P pushed up to within a point of new all-time highs during the day but VIX futures diverged as can be observed in the intraday SPY arbitrage model.

With the VIX futures term structure mostly stationary the spread for the front two months remained at -0.85 making for a roll yield that isn't benefiting XIV much (this lack of movement has also resulted in a mostly stationary VXX Daily Forecast).

Spot VIX also diverged from its normal inverse correlation to the SPY again, making it 4 days in a row or positive correlation. I'd love to see someone run through the data on this to see when the last time was that this happened (typically positive SPY-VIX correlations are negative for the market in the following days).

VIX remains 5.6% below actual market volatility over the past 30 days (HV21 at 14.53) resulting in a continuation of a negative risk premium. While this is unusual it's not unheard of, especially after a recent spike in VIX like we saw in mid-April. If we get a few more low volatility days in the market HV21 will come down to about 13.75 by Thursday.

The daily SPY arbitrage model is still holding a pretty wide spread as well:

Given that the usual correlations seem to be temporarily broken and the contango spread is neutral it seems best to continue to wait it out a bit for a more profitable setup. Alternatively, if I owned XIV/SVXY I still think it's a good idea to pick up some cheap VXX calls as I mentioned via Twitter last Wednesday.


UPDATE
Performance of S&P 500 after 3 or more consecutive days of positive SPY-VIX correlation, from 3/2004 to present:

Retirement of the Twitter Sentiment Analysis Tool

While I find the information from the Twitter Sentiment Analysis tool to be interesting, it doesn't lead to any actionable trading in its current format.  Because of this I've decided to remove it from the site.

I will continue to track the sentiment data to see how it correlates with the price of VXX over a longer period of time to see if there is a more usable format, but I tend to think that the tool is better suited for gauging interest in individual stocks rather than ETPs.

As a final chart, below is a view of the daily net Twitter sentiment of VXX compared to the daily percent change of VXX over the past month.

The #1 Rule For Trading Volatility ETFs

Every couple weeks or so I see a new article from someone giving advice on how to trade volatility ETFs & ETNs such as XIV, VXX, SVXU, UVXY, and ZIV. Often the author will at least know that these products trade based on the VIX futures. But every once in awhile I'll see something more misguided, like the chart below. While it looks promising, once it is applied to historical data it does not deliver positive results.

For example, let's take a look at the period between August 2004 and May 2006 when VIX bounced between 10 and 16 (VIX on left axis; VXX on right axis). The chart says buy VXX below a VIX of 16.

The return for VXX during this time was -75%, a loss attributed to the fact that the front two months were in contango with a spread averaging 0.98.

Let's continue on to the next period, between May 2006 and Feb 2007 when VIX was spending much of its time between 10 and 12.  The result for VXX is a 57% loss, and again the reason is a contango term structure, with an average spread between front month futures (M1) and 2nd month futures (M2) of 0.85.

You may point out the spike in VXX from $1000 to $1350 on the left hand of the chart, which is a nice 35% gain. Unfortunately, if you were following the chart and buying XIV when VIX rose above 16 then you would have lost 30% over the next couple of weeks before almost breaking even if you held your position until VIX dropped below 16 again.

OK, so I can hear some of you saying that I'm cutting off the chart before VIX and VXX spikes. Take a look at the next period, from Feb 2007 to Feb 2011 when VIX spiked during the 2008 crash. During this time VIX was above 16. Since the chart above says buy XIV when VIX is above 16, I'll compare VIX (left axis) to the price of XIV (right axis).

The result is a 64% loss up until Dec 2007, followed by another 66% loss from Aug 2008 to Dec 2008, for a total of an 88% loss in XIV from 2/27/07 to 12/19/08.  At this point your position has essentially been demolished after this 21 month period. But let's give you the benefit of the doubt and say that you stayed solvent and held on to this position for the 4 year period until VIX dropped below 16 again. In that case you managed to almost break even in that XIV trade.

Looking at the term structure during this period, we find that between 2/27/07 and the bottom in 12/19/08 the term structure was in backwardation, with an average of a 0.67 spread between the first two months. Backwardation has the effect of making VXX rise and XIV fall.

From 12/20/08 to 3/24/2011, when the price in XIV goes up from 2 to 14 for an excellent 7x return, the term structure is back in contango with an average of 1.51 points between M1 and M2. Contango has the effect of making XIV rise and VXX fall.

At this point, the #1 rule in trading volatility-based ETFs should be clear: Follow The Term Structure.

The term structure is going to be the biggest driver of price for volatility-based ETPs over time. There are of course other factors that need to be taken into account when trading volatility ETPs in order to maximize your gains, but you will be able to trade these products pretty well just by following this rule.

To view the current term structure and other metrics that are critical to trading volatility ETPs, visit the VIX Futures Data page.


**Note: VXX and XIV price data prior to the funds' first trading days (1/30/2009 for VXX and 11/30/2011 for XIV) have been calculated by using the VIX short-term futures underlying index values derived from actual data of month 1 and month 2 futures.

How The SPY Arbitrage Model Can Be Used In Trading

In March I announced the availability of the Intraday SPY Arbitrage Model.  This tool provides traders with information about how the S&P 500 is tracking against other asset classes during the day and can be helpful in various ways. Today I'll discuss two ways of using the arbitrage model to place market-neutral arbitrage trades.

For reference (and perhaps for those with short attention spans), I'll provide the weights up front. For the curious skeptics I'll explain below.

SPY:  +$12
XIV:     -$1
TBT:     -$2
HYG:  -$13.2

The first way is to trade using the intraday SPY arbitrage model. Below is a graph from March 8, 2013 showing the difference between the price of SPY and the model:

In order to take advantage of the arbitrage opportunities such as the $0.40 spread in the last hour of trading, one must know the correct weighting of securities. With the weights applied correctly the result would be to capture any difference between the SPY and model from the time you enter to the time you close out the trades.

In what is seemingly a ridiculous chart, below is a comparison of the the above chart (difference between the SPY and the model) and a trade consisting of the components and weight ratios as specified above. The point here is that they are identical, thus proving that by using the weights specified you get the exact same result as the difference between the SPY and the model.

In this particular instance, since SPY is higher than the model a trader would short the SPY and be long the model components in the weights given. Specifically, short $155 of SPY and long $12.92 XIV, $25.83 TBT, and $170.5 HYG (these numbers come from the -12 : 1 : 2 : 13.2 dollar ratio).

Side note: If you can not find shares to short for a given security, you can substitute in the inverse ETF (e.g. long VXX in place of short XIV, or long SH in place of short SPY) using the same dollar amount. You can also reduce the amount of capital required by substituting in a leveraged ETF (e.g. long UVXY in place of long VXX, or long SSO/UPRO in place of SPY), but be sure to reduce the amount of dollars for that component by half (or by a third if you are using a +/-3x ETF in place of SPY).  The net result will be the same with these substitutions.

Assuming the standard securities, the example above requires a total short position of $155 and total long of $209.25. Your gain on this low-risk trade, should the spread collapse from $0.40 to $0.00, would be $0.40.  If the spread increases further the trade would lose money for each penny it widens. Note that the spread does not always approach zero at the end of the day, and intraday spreads as large as $0.80 or higher can occur.

For a more practical trading scenario, multiply the standard dollar amounts by 100 for a short position of $15,500 and total long of $20,925.  The net gain on a spread collapse from $0.40 to $0.00 would be $40. As mentioned above you can reduce your capital requirements by substituting in leveraged ETFs, but this trade clearly requires a large amount of capital and low commissions to be of much use. (**For potential issues using this trade please see the end of this post.)


Daily SPY Arbitrage Model
For traders looking for a longer term play, the same weights can be applied to the Daily SPY Arbitrage Model which updates prices only at the end of each day.  In this model a trader can place trades spanning weeks or months and take advantage of larger arbitrage spreads. For example, the current spread is nearly $10 as shown in the graphs below. Using the practical trading example (discussed above) of $15,500 short and $20,925 short, a collapse of a $10 spread to $0 represents a gain of $1,000.

Knowing at what size spread to open a trade can be difficult in this model and requires some analysis of the bond market in addition to stocks. Given that central banks around the world are doing all sorts of things to influence bond prices, prices in different markets can be disconnected for an extended period of time. These factors make this trade suitable for advanced investors only.

One big difference in the daily model vs the intraday model is that if you take long positions in leveraged ETFs for a trade and hold for any duration of time longer than a day you will open yourself up to negative compounding errors. I won't go into the details on that concept but you should do a Google search on it and know what it is. To turn the compounding errors around in your favor, you can take short positions in the inverse leveraged ETF for the position that you want to take. For example, go short SPXU with a 1/3 dollar amount instead of being long SPY.


One last point of clarification for both models is that all the arbitrage models charts currently reference TBF as the treasury component. The intraday model actually uses TBT for calculations because it has a higher trade volume than TBF. The daily model obtains data using TLT in order to avoid distortion of the model caused by compounding errors. Regardless, the model weights use TBT as a reference because I find it to be easiest to communicate the weights with consistency. Feel free to use your favorite 20+ yr treasury ETF but be sure to make adjustments off of the TBT weight.




**Some issues with using the model:
- You may run into problem finding shares available to short through your broker, especially on lower volume securities. This can be a problem specifically with HYG which does not have an inverse ETF.
- If you currently have an open short position, your broker could request that you close your position (or force close it ) if they get low on available shares.
- If you are subject to wash sale rules your losses may not properly offset your gains.
- The model is based on several indexes and will only be accurate to the extent that they are tracking their intrinsic value.  This can be checked on your trading platform.  For example to check XIV you would look up the symbol "XIV-XIV.IV".
- When there are gaps up/down in the morning there is the possibility that the intraday data doesn't line up correctly and needs to be adjusted.  If the arbitrage opportunity looks too good to be true (especially early in the morning), it probably is (check the SPY vs Model Components graph to make sure they are all starting together).

Daily Wrap and VXX Forecast For 4/19

These are difficult times to trade volatility with a high degree of certainty, given that the range of VIX and 7 months of VIX futures is only 2.05 points (17.15 to 19.20). The compression represents a lack of direction and takes away the ability of traders to use the roll yield of VIX ETPs to their advantage. This can be observed on the right hand side of the chart showing 7 months of VIX futures, below (from VIX Futures Data page):

I noted yesterday that "there is a good probability that [the VIX] will retrace some of those gains tomorrow but by no means does it have to, and a spike today does not rule out the possibility of another spike tomorrow." After today's action I think the same thing can be said for Friday's outlook as well. Staying in cash until we get through this choppiness can be a great decision.

VIX has closed above front month futures for 4 days in a row now so risk remains elevated. Forecast for 4/19:

Daily Wrap and VXX Forecast For 4/18

In yesterday's post I noted that the risk of a spike in VXX was quite elevated, with the needle on the VXX Spike Risk gauge approaching the red zone. Today VXX closed at 21.07 after a moderate/strong gain of +11.5%.

The term structure flattened out a bit but is still in a slight contango, producing nearly irrelevant roll yields for XIV and ZIV of +0.5% and +1.65%, respectively.




Spot VIX closed 0.1 above front month VIX futures at 16.51, making a short of VXX very risky. Given the 18% spike in VIX today there is a good probability that it will retrace some of those gains tomorrow but by no means does it have to, and a spike today does not rule out the possibility of another spike tomorrow. Generally these are the types of situations where I prefer to wait for a setup that is more predictable.

The VXX forecast for tomorrow (preview below) generally reflects a neutral bias and a reduced, but still elevated, risk of a VXX spike.

Daily Wrap, April VIX Expiration, and VXX Forecast for 4/17

The term structure reverted back to a (slight) contango today with April and May separated by 0.8 points, as seen in the term structure from the VIX Futures Data page.

Since VIX futures roll tomorrow morning I care more about the May and June futures, which are separated by 0.95 points.  This will set up to be a good chance to buy XIV if the market can find any foothold tomorrow. One problem with the trade, however, is that VIX is not really overpriced here. Historical (actual) volatility over the past month is now 13.47, with VIX just 3.7% higher, as shown in the HV vs IV chart:

Sellers of options want to collect a premium that is reflective of historical conditions and future risks and it remains to be seen if traders think that the action in the past couple days was just a fluke. If we can get some calm days in the market, both VIX and HV21 will drift lower, possibly as low as actual volatility over the past 3 months (HV63) which is currently around 11.0 -- ~20% lower than where VIX is currently.

The other risk to the trade is that May futures aren't really overpriced either. At 14.75, May VIX futures are just 5.6% above spot VIX, meaning there is not much buffer to absorb a spike in VIX, making a sharp drop in XIV very possible if the market sees more selling.

In addition to the roll yield of about 1.4% per week on XIV, another positive data point for the trade is that the market is down 1.2% over the past three days while VIX is up 14%. This suggests that the move in VIX may have been overdone, especially if you think the market has already priced in risk from what has happened over the past few days..

Taking a look at the the VXX Forecast for tomorrow (preview), we can see that the roll yield and risk of a VXX spike reflect the numbers discussed above.

XIV: When A "Sure Thing" Goes Bad

A shift occurred in the market today with 1st and 2nd month VIX futures closing in a very slight backwardation of 0.05 points, with April at 16.65 and May at 16.60 (see data page). While this does not necessarily mean a continued selloff in XIV, I think it is very important for anyone who has grown accustomed to the seemingly unidirectional movement in XIV to understand that gains are not always a given.

We've seen a good run -- the result of a VIX that has declined from 26.66 on 6/1/12 down to six-year lows of 11.03 in March, as well as rolling futures in a contango term structure.  The result has been a staggering 200% gain over the past 9 months, and a 400% gain since 11/21/2011. It seems that money has been raining from the skies!

But remember that key drivers of VIX futures ETFs are 1) change in price of the relevant VIX futures contracts and 2) the term structure. Today we saw a shift in the flattening of the term structure which should signal a very loud "caution" to you.

When it comes to trading VIX futures ETPs it is critical to watch  for these changes in market structure.  Consider the following graph of XIV from 2004 through 2013 (prices prior to the fund launch have been by calculating the index value using 1st and 2nd month VIX futures data).

Those are some seriously painful losses! The thing that all of those declines have in common is a term structure shifting into backwardation. This is why it is so important to watch the term structure -- something that I watch daily by looking at the VIX Futures Data page (see 'M1-M2' for XIV and 'M4-M7' for ZIV) or the Daily Forecast (see the 'VXX Bias' gauge).

Exiting long XIV positions when the term structure flattens before a move to backwardation, can help you avoid these losses.



Today's closing term structure:

Tomorrow's VXX forecast: no bias from roll yield.  Spike risk remains elevated.:

Weekly Wrap 4/12: Easy Money

Pretty nice week for those owning XIV with a 8.8% gain.  Term structure and VIX declined substantially and remain in contango producing a roll yield over 2% per week for XIV and almost 3% for ZIV (see VIX Futures Data page).  View of the term structure week over week:

Looking Ahead:
Next Wednesday April VIX futures expire and all months roll forward. You can expect April VIX and spot VIX converge as we get closer to Wednesday.  Right now April is ~5% higher than VIX so there's some  premium left but not much. Don't expect disproportionate downward moves from VXX this time around.

We've been in a pattern in the market lately where it is ridiculously easy to make thousands of dollars just buying the dips of XIV in the morning and selling during the afternoon melt-up. Cash out and repeat the next day. Seems to be a common occurrence and everyone is on to it, making me wonder just how much longer that pattern will last. This is trend we are currently in and I choose to play along rather than try to short it, but I remain very cautious with what I hold overnight.

Checking in on the VXX Daily Forecast for Monday, we can see that it has changed very little over the past week due to a pretty consistent contango term structure and positioning of front month futures relative to VIX. The bias remains slightly negative after today, however the risk of a spike edged up a bit.

Weekly VIX ETF Performances vs S&P 500:
XIV:     +8.8%
ZIV:      +5.2%
VXX:    -9.2%
UVXY:  -17%
SPY:     +2.3%

Actual Volatility and Forward Implied Volatility Continue To Diverge

Actual historical volatility for the S&P 500 over the past 3 months (HV63) fell to 9.24 today, putting current 30-day forward volatility (VIX) at a 50% premium. From the VIX Futures Data page:

This is getting to be a pretty large gap and it looks like it could be a good time for new long positions in XIV in the next day or two. However the risk of a position in XIV right now is that the premium between VIX and front month futures (the yellow and blue lines above) is only 2%. This means that if we do see a VIX spike there is very little "buffer" in M1 to absorb the spike so it will be more likely to see gains as well, especially if VIX stays above M1 for a few days.

We can see this risk reflected in the VXX Daily Forecast gauges. A short VXX (or long XIV) position is still in favor (just barely), but the risk of a spike has been increasing over the past several days. In fact, if you look at the daily chart of VIX you'll see that it's been on a choppy rise over the past 2 weeks -- a pattern that sometimes leads to a large VIX spike. I still think that a smaller position or no position is justified until we see some real relief in the VIX..

Some other interesting action today can be observed by using the intraday SPY arbitrage model, which seemed to be all over the map.

Short term futures were up over 2% (as seen in the decline of XIV), pricing in a downward move in the SPY. The term structure for the first two months flattened to under 1 point until about mid-day when XIV decided to reverse to catch up to SPY and close up 1.5%. VIX futures closed lower and with a wider contango spread (-1.3).

Treasury yields decoupled from SPY, falling all day and closing substantially lower (see TBF). 

High Yield Credit (HYG) sold off pretty hard toward the end of the day and finished negative.

So a bit of disagreement between assets and reason for continued caution. 

New Tool Available: VXX Daily Forecast

I provide quite a bit of essential data for trading volatility ETPs on the VIX Futures Data page, but if you are unfamiliar with the various VIX metrics it can be difficult to know where to look and how to interpret that information.

For me, the numbers on the page are critical since I look at them at the close of each day to determine whether I want to buy, sell, or hold various VIX ETFs. Since I often receive questions about whether it is time to be long/short XIV/VXX/UVXY, I decided to share a daily forecast tool for VXX which essentially summarizes my opinions on the direction of VXX for the next day of trading.

There are two components to this forecast, as illustrated by the gauges on the Daily Forecast page.

1) VXX Bias 
This gauge measures the bias of VXX and is largely determined by the roll yield (more on the roll yield in this post), which can be conceptualized as a headwind or tailwind. The gauge values are on a linear scale from -10 to +10, with a more negative reading indicating a stronger negative bias (headwind), and a more positive reading indicating a stronger positive bias (tailwind). Note that the gauge's range covers +3 and -3 standard deviations from the past 9 years of data for a total of 6σ.

2) VXX Spike Risk
The second forecast gauge reflects the likelihood of a significant spike in VXX. It takes data from the term structure of the front two months of VIX futures, the premium of front month futures over (or under) spot VIX, and recent moves in spot VIX. The gauge output is a scale of 0 to 10, with 0 being the lowest risk of a VXX spike, and 10 being the highest risk of a VXX spike.


Using the Gauges
Shorting VXX (or being long XIV) has been an excellent trade over the past several years due to a persistent state of contango in the VIX futures term structure. Simply being short VXX when there is a negative bias and long VXX when there is a positive bias has resulted in an average return of 116% per year over the past three years. The VXX Bias gauge will tell you which side of the trade you want to be on. Values toward the middle, and specifically in the yellow zone, mean that the term structure is relatively flat and you won't get much benefit from the roll yield with either a long or short position. It's important to note that while the bias gauge provides insight on the headwind/tailwind that VXX faces, VXX can still move up with a negative bias or down with a positive bias if the underlying futures move enough.

When you are short VXX it is always important to know what the risk is of the trade moving against you. While VXX loosely follows the daily action of the VIX, future VXX performance can be determined by the shape of the term structure in the 1st two months of VIX futures, and the value of these futures in relation to spot VIX. The readings on the gauge will spend a majority of time in the yellow region, signifying a medium risk of a spike. A move into the upper range of the yellow and red will make me increasingly nervous to be short VXX or long XIV.

Of course there are market/world events which will cause the VIX and VXX to spike dramatically. While this model cannot predict the unexpected, it does provide the ability to detect shifts in investor sentiment of known macro risks.

Below are some examples of what the gauge look like under various conditions (it might be helpful to look at the charts for VXX/XIV to see what price was doing during these times).

First is a series of five dates showing the period from late July/early August as the term structure flipped from contango to backwardation (some days omitted due to lack of substantial difference).

_________________________________


July 21, 2011
M1-M2: -1.65
M1 to VIX: 4%

VXX: $83.52


_________________________________



July 26, 2011
M1-M2: -1.15
M1 to VIX: -3%

VXX: $88.00


_________________________________



July 27, 2011
M1-M2: -0.40
M1 to VIX: -7%

VXX: $93.20


_________________________________

July 29, 2011
M1-M2: 0
M1 to VIX: -16%

VXX: $93.64



_________________________________

Aug 5, 2011

M1-M2: +2.95
M1 to VIX: -9%

VXX: $121.24


_________________________________________________


The next pair show a period of contango with a very high premium of M1 to VIX

_________________________________
Feb 17, 2012

M1-M2: -2.15
M1 to VIX: +27%

_________________________________________________


Lastly, a view of a very large contango spread:

_________________________________
March 16, 2012

M1-M2: -5.45
M1 to VIX: +12%