One of my favorite quotes was from John Sculley, a former executive with Apple and Pepsi. The quote was “Perspective is worth 50 IQ points.” As I get older and hopefully wiser, I find it hard to know if my perspective is the right one, so I never know if I’ve added or subtracted those 50 points. But I’m pretty sure that the current hyperbole in the business media is misplaced, and its effect on trader psychology is reflected in frenetic, go-nowhere market action. I’ll explain in a small rant.

My Small Rant

On August 23, China announced retaliatory tariffs of $7.5 billion, and the world turned a bit upside down. Put that $7.5 billion in some perspective. The $7.5 billion amounts to a 0.3% average price increase on total exports of $2.5 trillion, and that is only true if there are no mitigating actions taken by market participants and currency exchange rates stay where they are. Meanwhile, since January 2018 the US dollar has increased in value by 9% when measured against a trade-weighted foreign currency basket. So for the last 20 months, while the economy has been thriving, the exchange rate imposed a 9% price increase on that same $2.5 trillion export portfolio. These are sober facts you’ll never hear on the agenda-laden, sky-is-falling business channels.

So looking at the macro picture for exports, the exchange rate imposed a $225 billion price increase on exports of goods and services and China tariffs will impose a $7.5 billion price increase. Therefore, following the market’s manic logic, the $225 billion did not eradicate 2% to 3% GDP growth, but the $7.5 billion should tank markets. Where is the perspective?

Oh wait! It wasn’t the tariffs; it was Trump’s tweet. A bit later that Friday Trump tweeted this, and suddenly the Dow was down 400 points:

“Our Country has lost, stupidly, Trillions of Dollars with China over many years. They have stolen our Intellectual Property at a rate of Hundreds of Billions of Dollars a year, & they want to continue. I won’t let that happen! We don’t need China and, frankly, would be far…. better off without them. The vast amounts of money made and stolen by China from the United States, year after year, for decades, will and must STOP. Our great American companies are hereby ordered to immediately start looking for an alternative to China . . .”

Can you imagine a responsible CEO who hasn’t set about assessing alternative supply chains? I can’t. Where was the cataclysmic news in that? That Trump was in a trade dispute with China or that companies should look for alternative supply chains? Again, where is the perspective?

And what ammunition has China got left? They could tax the remaining $45 billion of our $120 billion total China exports, but that would probably not target Trump supporters (They’re real goal is to get the tough guy out.). The $75 billion represents exported goods, and the $120 billion is a combination of both goods and services. So, more likely a ratchet up would target consultants and other service providers, and how much intellectual property would become unavailable to China if they did that? Rant over!

Markets and Risk Inference

Anyway, the ETF basket I track (SPY, DIA, QQQ, IWM, and SSO) dropped 3.0% that Friday. Risk off!

Here is how VIX futures responded to that Friday’s chaos. Movement toward backwardation implies increasing risk.

VIX Futures Curve Progression, Friday, August 23

Source: Michael Gettings Data Source: CBOE

So, then came Monday, the 26th, and I posted this in a blog: “The markets opened up this morning and then began to fade . . . At about 11:00 EDT, with the ETF basket still up about 1.0% the algorithm indicated a “Sell” signal.” The rapid move toward backwardation shown in the chart, and more importantly the rate of change, had triggered that sell signal, so I sold the ETF basket and bought IEF (10-yr treasury ETF). If you’re unfamiliar with the Easy VIX algorithm, I’ll explain much more later.

Meanwhile, treasuries were holding a strong rally – up 0.7% on Friday the 23rd and 1.1% by Wednesday.

I won’t bore you with all the price action at the end of the week other than to say the yield curve inverted further and the average ETF-price change, up or down, was 1.4% per day for the 5 days from Friday to Thursday.

By week’s end, the ETF basket composite price was $265.54, down $.08 from its $265.62 closing value of Thursday August 22 before the chaos. The net decline was 0.03%, that is three hundredths of a percent – a photo finish. So, in the end, the China retaliation, Trump’s tweet and the inverted yield curve meant nothing to equity prices. But somehow it didn’t feel so peaceful. And my algorithm called a “buy” or “reentry” on Friday August 30, another short sell interval with no follow through and a small lost opportunity.

In a blog last week, I wrote that the current market action is “like a car rocking in a deep ice rut, back and forth and back again, but going nowhere – going nowhere with a vengeance.” The Easy VIX algorithm is a risk mitigation tool, and it earns superior returns when markets follow through on trend reversals. Frenetic inflections every other day just confuse it.

To put a picture to the confusion, consider the shape of alternative VIX futures curves during recent risky and safe market periods. Risky curves are typically backwardated, higher values in near months. Safe curves are typically contango, lower values in near months. Transitions are normally pretty orderly.

Typical Risky and Safe VIX Futures Curves

Source: Michael Gettings Data Source: CBOE

In contrast, here is a graph of the VIX futures curve at close of business Friday August 30; it’s not so orderly

VIX Futures Curve, August 30 Close

Source: Michael Gettings Data Source: CBOE

As I said, the current market hyperbole confuses me and it’s confusing the VIX futures curve as well.

Why It Matters.

The Easy VIX algorithm extracts risk-mitigation signals by measuring the rate of change in the daily shape of VIX futures, and when triggered, I sell the broad-based equity ETF basket and buy a 10-year treasury ETF, IEF. The algorithm has called an average of 6 sell intervals per year since May of 2008. Seven of eight sell intervals result in a zero-sum distribution of small advantages and disadvantages, but one‑in‑eight sell intervals avoid large losses and facilitate reentry at bargain prices. Those signals produce returns that are about double those of a buy-and-hold strategy and, at least as importantly, significant drawdown protection.

For those who care, at the end of this article, in “The Quant Corner,” I’ll discuss some of the modeling parameters and some improvements I’ve made recently. For now, I want to show how the frenetic market behavior is reflected in the algorithm, and what implications might be drawn. Here is a graphic of the algorithm’s performance, including the new refinements, compared to holding the ETF basket over the last 12 months.

The Easy VIX Algorithm Performance August 2018 – August 2019

Source: Michael Gettings Data Sources: Fidelity,, CBOE

Notice that the performance largely stems from the December 2018 correction which provided an opportunity to ride treasuries through the dip, reestablish positions in the equity ETF basket at lower prices, and then compound subsequent growth on the higher investment base.

But my focus right now is not the gain profile, it is the number of sell intervals. Since 2008, the model produced about 7 sell intervals per year; in contrast, there have been 16 intervals over the last 12 months. (The improvements make the model a bit more active than the earlier version.) Recently the risk-on/risk-off cycling has become manic, and the current frequency of sell intervals looks a lot like the frequency leading into December 2018. Viewed another way, markets were healthy from early January through the end of April, and very few sell intervals are seen in that period. But the frequency of signals has become increasingly rapid in the last two months, much like the prelude to the December correction. There have been so few corrections in recent years that I don’t have any statistical reference as to what pattern warns of a major downturn, but the pattern I see here does make me cautious.

So, what do we do in this environment? Rapid and erratic price moves with no follow through in either direction make managing risk a chore. Selling into a risk-off period that follows through with a big correction is truly rewarding when we reenter at bargain prices but selling into a risk-off period that then evaporates, just to buy back at similar price levels is tedious. Yet to date I have no way to separate the tedious noise from the big downturns. What I do know is that missed opportunities will be small and they will balance out with small advantages, but about one-in-eight big moves will be very rewarding. So, I go about the business of exiting on sell signals and hoping for the big downturn that enables a big profit on reentry.

The Quant Corner

I always struggle with how much of the quantitative side to share. On the one hand, it builds confidence for those who care; on the other, it will make some readers’ hair hurt. For the most part, I’ve gotten over my concern that someone might reverse engineer the system since the AI algorithm provides a proprietary firewall.

So, I came up with an idea to create an optional section – The Quant Corner. If you care to know more, read on, otherwise skip to the end.

I’ve built two refinements into the algorithm aimed at addressing its propensity to be a bit late in reentry calls, thereby foregoing some advantage that could have been captured with a slightly earlier buy signal. The earlier graph of the last 12 months’ performance reflects the changes. I’ll discuss them, but first I’ll outline a few basics to set a foundation.

The Easy VIX algorithm was designed as a risk-mitigation tool. It also happens to produce superior returns. There are three principle metrics that, together, identify sell and reentry signals. All of them are derived from daily measurement of the contango or backwardated shape of the VIX futures curve as shown in the graphs of the main article above. There are three possible +1-point values and one possible negative 1-point value; an aggregate score of +2 signals ‘sell’. Anything less indicates a ‘buy or hold’. This key explains the basics.

When a “sell” is triggered, I sell the broad-based equity ETF basket and buy 10-year treasuries (IEF); I sell IEF and buy the ETF basket back when the buy-or-hold signal returns.

The two new model refinements are these:

  1. For reentry decisions only, the Primary Slope now reflects an exponentially smoothed one-day forecast. The Primary Slope forecast is derived from observed Primary Slope metrics where the artificial intelligence algorithm resets the look-back horizon periodically based on then-available historical data, applying calibrations prospectively once set.
  2. The other refinement is that now the Primary Slope look-back horizon is adjusted in accordance with observed price volatility, measured over a trailing 30-trading-day period. So now higher volatility drives shorter look-back horizons. The effect is to make the Primary Slope more responsive when market movements become more volatile. This is especially true of recent price action.

Combined, the changes add about 2% to the average annual returns over the eleven-year period bringing the price-only internal rate of return to 17.5% versus 9.0% for the reference buy‑and‑hold strategy. Drawdown protection also improves. Using the algorithm, no loss occurred when smoothed over rolling-12-month periods. However transient losses occur over shorter periods given lack of recovery time. The following table shows the drawdown characteristics and the effects of recovery time for rolling periods of different terms:

Drawdown Protection for Varying Terms(252 trading days equals one year)


B&H Worst G/L

Algorithm Worst G/L

Drawdown Advantage

252 days




126 days




63 days




The drawdown matrix makes a point that I’d like to emphasize. No risk mitigation protocol is perfect, and when looking at outliers the shorter the horizon the less effective it will look. Said another way, if anything is 90% effective, inspecting smaller and smaller intervals will eventually show worst-case outcomes that fall into the 10% ineffective zone. Over longer periods, the ineffective results get overshadowed by the dominant effective results. Since I choose to look at risk-mitigation effectiveness in terms of worst-case drawdowns, I’m not at all surprised to see the numbers in the chart above. Keep all this in mind when a given 10-day sell interval turns out 2% worse than holding for the period.

Having said that, the refinements do make the algorithm more responsive at times when cycles are short-lived. The current environment is like that, at least for a while. Here is a graph comparing the new algorithm to a buy-and-hold strategy over the last year.

Comparison of Refined Algorithm v. Holding, Last 12 Months

Source: Michael Gettings Data Sources: Fidelity, VIXCentral, CBOE

Just like the original version, the model performs very well in December, but it also does a better job of extracting value from the early-August price dip. You might recall that the early-August sell interval had accrued an advantage in excess of 3.1% by August 16 only to watch it dissipate to a 1.2% advantage by August 19 when the model called a ‘Buy’. Those results were published in Seeking Alpha articles here – August 16 article and August 19 article. That particular outcome is what caused me to search for a solution to the delayed reentry problem. If you inspect the graph during the August period, you’ll see that the new model gives up little value in August while the hold strategy dips materially.

So, from here forward I’ll be using the new Easy VIX algorithm. I’ll also observe version control. Any exploration of further improvements will be done in a development environment, while “production” signals will use this version until further notice.


I hope this brought you some useful information and maybe some useful perspective. Right now, I’m invested in the ETF basket. I monitor the metrics every day and have taken to posting frequent blogs when I see something worth reporting. Articles like this are subject to editorial review before publication, so they’re often not timely enough to convey trading signals. If you’d like more timely market signals, please search for my blog posts a couple of times each week; no notices are sent about blogs.

If you’re not a follower, please click the orange button to the right of the articles title, and feel free to post comments. I enjoy the exchange of views. If you are already a follower, thank you.

Disclosure: I am/we are long SSO. 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 trade all the tickers mentioned using the algorithm described. The artificial intelligence algorithm monitors daily performance and periodically recalibrates look-back horizons and triggers in a step-wise sequence. New calibrations are applied prospectively only, and never applied to the historical period from which they derived. The algorithm described and the discussions herein are intended to provide a perspective on the probability of outcomes based on historical performance. Neither modeled performance nor past performance are any guarantee of future results.

Source link