A Barbell Portfolio Strategy

with a realtime review of portfolio performance during the recent market pullback.

I want to take a look at some of my core portfolio strategies from a fresh angle. The basic principles discussed are easily transferable to other types of portfolios.

A key portfolio feature, I strive for, is an adaptive combination of strategies with distinctly different return characteristics. These strategies perform well in different market environments and, as long as these regimes are well recognizable and stick around for some time, strategy exposure can be adapted to the market environment. Such a portfolio should theoretically be able to achieve positive returns at virtually all times – though in real life sometimes even unrelated ideas will start moving in lock step. Losses can never be avoided, we can merely aim to keep drawdowns shallow.

I see such a holistic view very rarely implemented in practice – most products and services are aimed at specific niches and many investors are biased towards one strategy or asset allocation over another, because of their underlying investment philosophy. For me the most promising approach is to diversify across a number of different approaches and philosophies that all work, with the probability for a positive outcome on their side – as long as I can reasonably include them in my portfolio and investment beliefs.
Every single portfolio element may underperform or even fail, but the differentiation against both a classic form of portfolio allocation and individual trading strategies lies in balancing the diverse parts. Only the result of the portfolio as a whole counts and over time this result should be superior to any single strategy or asset allocation on its own.
In order to succeed, I frequently analyze the role each component plays in the whole, especially when special market events stress-test them in live conditions. Awareness of each element´s characteristics, dangers and likely performance in different market regimes is immensely important.

At the beginning of February 2018 a rather extreme equity market pullback (especially when viewed through the lens of volatility) happened. While it was overdue, it also arrived quite suddenly with few and subtle indications. This quick shift from the low volatility extreme in 2017, with record strings of new equity highs worldwide combined with unprecedented volatility lows across asset classes, to the most extreme volatility spike ever seen in the VIX (up 115.6% on February 5th alone), was especially lethal to popular short volatility strategies.

As short volatility strategies are a core component of what I invest in and describe in this blog, a good question to ask would be: Well, how did the portfolio do?
I´ll dig into a detailed review of this question, but let´s first look at the underlying investment philosophy.

The Barbell Idea on Steroids
I will concentrate on analyzing the satellite strategies around my core portfolio, which consists of a global asset allocation with a few twists (factor exposure and adaptation to market regimes) – this is a strategy many market participants use: it will go up in-line with the global market of all publicly investible assets (hopefully a bit more) and down with it as well (though maybe a bit less).
To provide a counterweight to extended drawdowns as well as boosting returns of the core asset allocation in good times, a barbell of satellite strategies comes into play.

The barbell idea is originally based on Nassim Nicholas Taleb´s notion of antifragility, where a small percentage of high risk investments takes advantage of extreme market movements, while the main portfolio is very low risk and ensures safety.
I put together another type of barbell portfolio, that includes strategies and assets that fall on opposite sides of the spectrum, selecting the strategies with the highest and most dependable risk-adjusted return from each end.
It balances two high risk, high expected return strategies that safeguard and complement each other because of their opposing return characteristics:

By default I equal weight my allocation across these two strategies and the core global asset allocation, but weights can change quickly as we will see in the realtime review below.
Both strategies have a high expected risk-adjusted return with historic Sharpe ratios around 1. They can be scaled to any desired level of risk, because both use derivatives and are easily leveraged at low cost – this is a great advantage, but also a danger that complicates implementation somewhat.
But the key to their superior portfolio contribution is, that they behave in opposite ways in almost any market condition. When one crashes and burns, the other usually excels. As both are expected to yield positive returns over time, this negative correlation will greatly reduce the overall risk and produce a more smoothly up-trending portfolio equity curve – this is the essence of the barbell idea.

Return Distribution
A major factor in the analysis of any strategy is its return distribution over time. Very few strategies have normally distributed returns, most are exposed to more extreme negative events, than a gaussian distribution would lead us to expect – which often leads to blow-ups, financial crisis and such. But some approaches are over-exposed to extremely positive events, a desirable quality, but in practice very difficult to implement and often paired with negative expected returns, eg. in the case of hedging.
These types of distributions are called negatively skewed and positively skewed, corresponding to convergent and divergent risk taking and they constitute the main opposing quality of our barbell:

Convergent Risk Taking leads to Negative Skew – e.g. a Short Volatility Strategy:
  • Many small gains are interspersed by occasional large losses. This is useful to produce a constant income stream, but downside protection needs to be very diligent as crash risk is high. Short volatility strategies, for example, have a strongly negative skew, but equity markets and other assets display negatively skewed returns as well.
  • Being short volatility creates reliable income in normal markets with sideways and regular up- and down-moves, it will underperform in strong up-trends and crash in strong down-trends, when volatility tends to shoot up – there needs to be a strict mechanism in place to exit the strategy, when volatility spikes, as well as diligent position sizing. The crash risk can stay hidden for years lulling investors into a false sense of security, as became obvious after the fact in the recent volatility event.
  • Negative skew strategies typically capitalize on mean reversion. Backtested Sharpe ratio is overestimated and needs to be corrected down. (I correct backtested Sharpe ratios of 1,2 to 1,4 down to below 1).

Divergent Risk Taking leads to Positive Skew – e.g. a Trend Following Strategy:
  • Many small losses are interspersed by occasional large gains. The problem when running, for example, a trend-following strategy is not so much a crash, but rather “bleeding to death by a thousand cuts“ – one has to be able to tolerate long strings of losses and often the bulk of the gains will be made by a single extremely positive position over a considerable time period.
  • Trend-following strategies are usually scaled by volatility and play both the long and short side of the market, adjusting to different trends and volatility environments automatically. They tend to outperform in the extreme environments of parabolic runs and epic crashes, while normal markets lead to underperformance and sideway moves to losses through frequent whipsaws.
  • Positive skew strategies typically capitalize on momentum. Backtested Sharpe ratio is underestimated. (I backtested and researched Sharpe ratios of 0,7 to 1 – a realistic risk-adjusted return in-line with short volatility strategies)

As both strategies have measures in place that reduce exposure in a negative market environment, they aim to be overweighted when they perform best and underweighted or stopped, when dangers are high and performance is poor.

Will this idea work out over time? A real-time stress test, as we experienced in the beginning of February 2018, is very valuable and should be examined closely.

Realtime Review
I’m writing this on February 10th after a roller coaster week with highly elevated volatility and the VIX Futures Term Structure in deep backwardation. It is unclear how long volatility will stay elevated and the further path of the correction is yet unknown.

On everybody’s mind and headlined at every news outlet is the epic crash of the short volatility complex, especially the crash and subsequent liquidation of short volatility ETN XIV. It went from a high $146,44 in January down to $7,35 on February 6th – a drop of about 95%, mostly in a single day.
The volatility spike and the magnitude of the daily drop in short volatility ETP was truly unprecedented, but by no means entirely unexpected – the possibility of XIV being terminated at VIX spikes over 80% is clearly stated in the product prospectus, which also says that its long-term expected value is zero. Apart from that a yearly gain of 180% for XIV in 2017, 9 times the S&P 500´s returns, was a common sense indication that high risk must be hidden in the ETN – after all return and risk are two sides of the same coin.
The crash was in large part caused by the hedging activity of short volatility products and traders themselves – the strategy had become overly popular and the volume traded in short volatility products was big enough to affect the market.
A ripple effect spread from VIX futures to equity markets in an inverse of their normal relationship. Directly after market close on February 6th the liquidity spiral, that began in the late afternoon, accelerated. This was triggered by short volatility ETP that mechanically hedged their exposure by buying large volumes of VIX futures at the settlement, pushing prices and VIX itself up, this spread into equity options and equity markets themselves, caused margin calls, position covering and so on.

The speed and the sheer magnitude of the unraveling of the short volatility trade was an impressive illustration of the tail risk vulnerability that a negatively skewed investment or trading strategy incorporates. Efficient protection must be in place to safeguard the downside in such strategies, but does that really work in practice?

So, let’s first look at the „death“ of short volatility, how it played out in my portfolio and at the question what this implies for the future of the strategy.

Theoretical Strategy Performance: Short Volatility
How did my short volatility strategy perform overall since the last entry signal on August 14th, 2017?
You can find the original strategy rules and detailed explanations here (part 2: Direct volatility trading) and an August 2017 review with some adjustments to those rules here – these rules are the basis of the performance analysis.
In general the strategy is in play when the Vix Futures Term Structure is in contango and the S&P trades above its 275 day moving average – when contango drops below -1,5% or the S&P falls below its moving average all positions are exited.
I use three different ways to implement the strategy, mainly to spread the risk across different entries and exits in case of just such an extreme volatility spike. I will also take into account that exit signals might have been missed. Let’s go through each scenario:

Long SVXY (XIV showed similar performance, but uses the slightly worse ETN structure, SVXY is an ETF and it is still tradable)
  • Entry at 9MA/60MA crossover on September 19th, 2017 with a position size of 10% of capital @ $89,73
  • Exit at close at the day of signal as per strategy rules: February 2nd, 2018 at contango -4,16% @ $105,16
  • Profit: 17% or 1,7% of portfolio value (NAV)
  • Alternative exit next day at close, when it became obvious that something bigger was afoot @ $71,82
  • Loss: -20% or -2% NAV
  • Or exit missed (as a worst case scenario) until the close on February 6th @ $12,24
  • Loss: -86% or -8,6% NAV

Long UVXY Put
  • Entry on August 14th, 2017, contango 5,82%, with a position size of 5% of capital @ $10
  • Exit two months later @ $15,4
  • Profit: 54% or 2,7% of portfolio value (NAV)
  • The volatility spike had no impact on this position, because of the time exit. However with a spike occurring within 2 months of purchasing the put, a maximum loss of -5% NAV would have been possible.

Short UVXY Call (I don’t trade this strategy, if it´s possible to miss the exit by a whole day!)
It is a little bit harder to be exact with the performance numbers here, because the data is hard to come by and I intermittently paused the strategy (see below). I extrapolate from the times I was invested according to the strategy rules, assuming a -200% stop loss for the whole position in February (an anticipated 3% NAV drawdown caused by the volatility spike)
  • Entry on August 14th, 2017, contango 5,82%, with a position size of 1,5% of capital
  • Rolling the position each time a 52% profit was reached (my average profit level of actual rolls)
  • Exit at -200% stop loss or at the latest at close of February 2nd, 2018 with contango signaling exit at -4,16%
  • Profit: 10% of portfolio value (NAV)

Overall performance: +14,4% portfolio value, 32% annualized.
Worst case scenario (missed exit in SVXY): +4,1% portfolio value, 8,9% annualized

Practical Strategy Performance
Of course reality is always a little different from theory, if you are not trading algorithmically. There are several concrete reasons for the differences:
  • Interactive Brokers raised margin requirements considerably in 2017 – these guys are not stupid and they perceived an increased tail risk at very low VIX levels. I had to reduce exposure which I did by selling SVXY almost immediately and UVXY Put positions early.
  • I stopped trading the strategies intermittently while traveling with uncertain internet access and over christmas – something I would have done with an algorithmic execution just the same.
  • In January I thought the S&P´s rise looked too good to be true, while observing rising volatility in a climbing market, which is an early warning sign. I decided to stay out of the short vol trade for the time being and wait for the next spike. In all I was completely out of the strategy in late December 2017 and the VIX spike had no effect at all due to dumb luck.
  • Returns more or less equaled out between theory and practice on a risk-adjusted basis (I spent less time in the market – making less profit, but avoided the volatility spike´s nerve-racking drawdown), but results would have been better overall, if I had just followed the strategy to the letter, as is usually the case.

This was the actual performance of the strategy in my portfolio:
  • SVXY: +0,15% NAV
  • UVXY Long Put: +2,57% NAV
  • UVXY Short Call: +8,3% NAV

Overall actual performance: +11,02% portfolio value, 24% annualized.

Even including the most massive VIX spike in history, the analyzed period was very positive for different short volatility strategies with a clear exit methodology – just about as good as it gets when compared to backtested results. The perfect low volatility environment of 2017 far outweighs the negative impact of the volatility crash. Drawdown was in line with expectations, because protective exit mechanisms were triggered well in advance of the crash. A moderate equity sell-off preceded the short volatility liquidation spiral, which was enough to start the exit signal flashing.
The probability is quite high, that in the future the exit signal will be triggered in time, but that is by no means a given – exogenous events (for example a sudden North Korean nuclear strike) may trigger a sell-off completely out of the blue.
Moderate position sizes are key to keep that risk from becoming potentially ruinous.

The Future of the Short Volatility Strategy
Not only do I think that it is possible to deal with the risks inherent in short volatility strategies responsibly, but also that the basic return drivers of the strategy are as valid as ever. Because of the returning fear in the market they are actually much stronger now and risk is lower then it was before the volatility spike.
Short volatility strategies are comparable to selling crash insurance for a premium. In the aftermath of a crash the available premium is especially high as demand for insurance is rising. Short volatility strategies usually recover quickly from their drawdowns when things normalize, because of these elevated premiums.
The high risk inherent in being short volatility is the very reason for the high return of the strategy in recent years and over long periods historically.

I`m not jumping back into the strategy just yet, though. Conditions are still negative and the true extent of the correction and duration of elevated VIX levels is unpredictable. The VIX Term Structure is in backwardation and short volatility positions are facing a strong headwind in the form of a negative roll yield, which has to be paid every day. Only when the Term Structure returns to its regular contango state and the S&P 500 holds above its long-term moving average, will the light flash green to start trading the strategy again.
Warnings abound, that many investment strategies, for example risk parity, are implicitly short volatility and may just be starting to unwind in the aftermath, causing the spiral to continue for some time.
Once a new entry signal is in place, I will stick to my position sizes and leg in slowly using different entries. I am thinking about substituting short volatility products by using VIX Futures directly to avoid being dependent on the providers of these products and paying unnecessary management fees to them.
Overall short volatility strategies are well suited to fulfill a specific function within a balanced portfolio, in my opinion.

The Performance of the Overall Portfolio specifically the Trend Strategy
Back to the barbell idea. The risky short volatility trade handled itself quite well all on its own, but what about the other parts of the portfolio?
I think, I was actually a bit overly discretionary in my portfolio decisions in January. Things seemed to good to be true, volatility was rising with rising equity prices and I started to raise cash and put on more hedges, then my research suggests is prudent. It was pure luck that these positions were hit by such a strong correction and more than compensated for any losses in the global asset allocation portfolio. I did also sell these hedges way too early and am now only protected by a large cash position reserved for resuming the short volatility trade.

The recently implemented Trend Following Managed Futures Strategy did well, but volatility was very high across the board causing nerve-racking equity swings above 10% of portfolio value – especially during Monday night (February 5th to 6th 2018). Many positions (eg commodities) were unaffected by the volatility spike and the main profit center of the strategy at the moment, being short treasury bonds, jumped right back to new lows after some wild gyrations. I will address lessons learned from this jump in portfolio volatility in a future post once the correction (?) is in the rear-view mirror.

Overall the portfolio passed the stress-test and, due more to luck than skill, actually is up more than 7% for the month in mid-February (see monthly performance stats here).
The real proof of the pudding is in the eating, though and only the performance in the next prolonged bear market will tell how well the barbell idea holds up under sustained adverse conditions.


Leave a Reply