Sornette's Supercomputer Is Betting On A Market Crash

Tyler Durden's picture


One of the world's most powerful supercomputers, retrofitted for trading the stock market, appears to be betting on a crash in the months ahead.

The Financial Crisis Observatory (FCO) at ETH Zurich released its latest Global Bubble Status Report on July 1st.

As we discussed with FCO’s director, Didier Sornette, on our podcast in May, they use one of the world’s leading supercomputers to monitor global markets each day for two distinct bubble-like characteristics: faster than exponential price movement and accelerating oscillations .

[ZH: We first discussed Sornette's work in 2010 on identifying "critical market crashes", and again in 2013 when Sornette explained "how we can predict the next financial crisis" to a TED audience...

The 2007-2008 crash seemed to come out of nowhere, with no source or group to take responsibility, an unpredictable one-time anomaly — as Sornette calls it: “the wrath of God.” But as he says firmly: Despite what standard risk management tools show, these outliers operate under special mechanisms that make them predictable, perhaps even controllable. Sornette and his team at the Financial Crisis Observatory (FCO) call these special cases “dragon-kings.” Dragon-kings, in direct contrast with “black swans,” are at the core characterized by a slow maturation of instability, which move toward a bubble, until the bubble reaches a climax and bursts.


There are many early warning signs of dragon-kings, but one of the crucial ones is super-exponential growth. Super-exponential growth is trenchant and unsustainable and can be found in many areas of study to predict dragon-kings. Sornette has applied it to Ariane rockets, parturition problems, epilepsy, landslides, even blockbuster movies and YouTube virality.


Dragon-king theory can be applied to 30 years of financial bubble history, starting with the worldwide bubble that started in 1980 and popped in 1987, and ending in the most recent global over-valuation bubble that broke in 2007 and 2008. In December 2007 Sornette predicted the Chinese market bubble, to the disbelief of analysts. Three weeks after his presentation the markets lost 20 percent, and by the end of the year they had lost 70 percent.


Can the dragons be slain? In a way. Learn the art of planning and predicting, says Sornette. If we find pockets of predictability, advanced diagnostics of crises are possible. So that crises may never again take us by such surprise.]

*  *  *

Sornette's July report notes an increasing trend of positive bubbles across multiple asset classes.

Here’s what they say in their “big picture” section:

“One can observe the continuation of a trend in the growth of positive bubbles in the fixed income asset class for the second month. The fraction of stocks diagnosed in a positive bubble state increased this month to exceed 36% compared with 32% last month. Mixed bubble signals still occur only in few commodity indices. We also observe renewed bubble activity in currency pairs.” [source]

Here is the chart where they show the “historical evolution of the fraction of assets within an asset class that show significant bubble signals”:

A positive bubble signal is an indication of herding when people start buying because prices go up. A negative bubble signal is an indication of herding when people start selling because prices go down.

Based on their daily scan of global markets, the Financial Crisis Observatory assigns individual stocks into four different quadrants based on their bubble strength and value score. These four quadrants are then used to create a trading strategy consisting of four different portfolios, which they define as follows (click image to enlarge):

In looking at each of the four portfolios, it appears that the supercomputer was mostly initiating long positions since March. However, starting in June, the researchers note that there has been a rebound in Contrarian Short portfolios in recent weeks:

“This month, we find that Long portfolios started to develop a drawdown in most portfolios initiated in March, April, May and June 2017. This reflects the stopping of the previously booming markets. Especially, the Contrarian Short portfolios rebounded in recent weeks. Contrarian Portfolios are more delicate to use due to their sensitivity to timing the expected reversal and exhibit very volatile performances, indicating that most of bubbles in the market are still dominating and that fundamentals have not yet played out. We expect trend-following positions to perform in the months following the position set-up and then contrarian positions to over-perform over longer time scales as the predicted corrections play out.”

Here is a chart illustrating each of their four portfolios since June with the rebound in Contrarian Shorts shown in green, which they expect to “over-perform over longer time scales as the predicted corrections play out.”

Based on the large and growing list of positions in its Contrarian Short portfolio—which includes the FANG stocks—it would appear that FCO's supercomputer is setting up for some sort of market crash or correction in the months ahead.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
The Black Bishop's picture

They need a supercomputer to figure that out? Jeezus....

38BWD22's picture



Did Martin Armstrong's supercomputer predict a crash as well?

Fahq Yuhaad's picture

Yeah, but it was his TRS 80 with the dot matrix printer.

The Real Tony's picture

Martin talks too slow and for too long. Most people would grow old and die of old age listening to him.

Singelguy's picture

Yes, he is expecting it next year.

christiangustafson's picture

The full-on panic wave 3 arrives a year from now.  First we get 1 down, and 2 to kissback the old channel.

Mtnrunnr's picture

try to time the short. I dare you.

logicalman's picture

If it didn't give a date, it ain't that super!

Uncle Sugar's picture


 What does the Gartman-1000 say?

A. Boaty's picture

My TRS-80 sez just the opposite.

rejected's picture

My Timex-Sinclare says that supercomputer is full of it!

Crypto Kevin's picture

Don't worry the PPT will save the day! BTFD!!!

agstacks's picture

I don't think so.  BOJ and SNB have "cash on the sidelines". 

MrBoompi's picture

I flipped a coin to see if there would be a crash.  I won the flip and you should be careful because sometimes I win coin flips.  

Arnold's picture

If you flip lands on edge, you get mind reading powers.
Saw it once, in the Twilight Zone.

Bob's picture

There was nothing hard about predicting the last crash when Michael Hudson did it in May, 2006.  It wasn't the least bit complicated:

He even included pictures, children's book style. 

Wrenching Away's picture

But its different this time!

syzygysus's picture

the Bernanke Ultra666 Quantum computer is front running all crash markers.  There will be no crash.

wholy1's picture

"Betting . . . "???  It's a G-I-V-E-N!

NoWayJose's picture

A Supercomputer prediction - just like SuperMan himself - always seems to come back to Earth!

halcyon's picture

No, the "computer" is not betting on a crash. Bad headline. Read the report here:

Yes, of course the crash will come, but not when the contrarian positions are piling in. Basic logic states that.

Trend following still looks to continue and to win, for the foreseeable months, until it doesn't work anymore.

Of course, something unexpected can come (and always does), and no amount of "supercomputer" can foresee that.


lester1's picture

Let's see here.. 

Sornette's Supercomputer vs the Federal Reserve's PPT


My money is on the unaudited PPT !

Fartboxbuffet's picture

Throw ya hands up in the air waive em like ya just dont care oooot ooot

JBilyj's picture

I just want to know how Gartman is going to bet...

RagaMuffin's picture

If the DOW runs to 23,000, is that covered by the supercomputer warranty? Or are the programmers drawn and  quartered ? 

Fahq Yuhaad's picture

This is pure bullshit babble to get more grant money. Count the hairs on your prick-- it's just as reliable.

medium giraffe's picture

Einstein's old institue, produced 21 nobel winners, invents supercomputer that predicts crash.

So we ARE fucked then.


Fahq Yuhaad's picture

Barffack Obongo was a Nobel winner too.

medium giraffe's picture

none of the recipients were awarded prizes for politicised nonsense.

Dragon HAwk's picture

To Really Err, you need a Super Computer.

But yeah Money for Nothing to buy Securities, wins every time.

Fahq Yuhaad's picture

The supercomputer will predict next week's weather as inaccurately as the conventional machine - it just does it a few million times faster.

trip kitchen's picture

It's a strange market you've created, Mr Yellen.  The only winning move is not to trade.  How about a nice game of chess?

InflammatoryResponse's picture

Bonus awarded for the Wargames Reference. :)


TrainReck's picture

Better known as squiggly line theory. I have a S&P crash indicator that has accurately forecast market corrections 45 % of the time (4 out of 9 Warnings since 2008) & the big crash in 2008. All but 1 Warning since 2008 has resulted in a sizeable pullback. Latest Warning signal came 6/30 the day the market was at 2451 ATH. S&P has been gyrating up and down since then, but 2451 looks like the top of the bubble. We'll see

Fahq Yuhaad's picture

Call it a "fully retro-tested proprietary oscillator" and charge $ 1000 per prediction.

Lore's picture

There's a radio show in Vancouver called "Money Talks" hosted by a guy named Michael Campbell where all the guests seem to have 20-20 hindsight: "The market could go up; it could go down."  Weeks later: "See?  SEE???  WE CALLED THAT."  Nauseating.  If they weren't in the financial markets, they'd make good "global warming" wonks.

Jim Shoesesta's picture

Pay me money to listen to more of my bullshit podcast, which, if were worth a shit, would pay for itself......

CoCosAB's picture



The day the 2nd Part of the GWT scheme begins!


In the mean time, slaves,

"Work, Consume, Obey"'s picture

"The 2007-2008 crash seemed to come out of nowhere..."


Hear that, Muppets?

Singelguy's picture

Really? I saw the crash coming in 2006 and sold all my real estate. Everyone thought I was nuts. To be honest, if I had held on for another year, I could have made a lot more, but as they say, pigs get slaughtered. There is definitely another crash coming. I am trying to figure out what the trigger is going to be. My best guess is a big bank collapse in Europe, most likely Italy.

In.Sip.ient's picture

Just what constitutes a "super computer" these days?

They used a CRAY for this???



TheMayor's picture

Most readers here have no idea what a Cray is.


Bananamerican's picture

but they know cray-cray markets when they see 'em

ali-ali-al-qomfri's picture

I'm going to need a bigger Abacus, mine is giving me rounding errors.

Yen Cross's picture

  Citadel= biis   (x) euroclear >  ex-im bank=  +/-   ecb* snb* (8) boj ^ _  Ag au ponzi  short<

RagnarRedux's picture

Super-Stupid Growth.

The Real Tony's picture

How can you have a correction in a long term bear market? That's a new one on me.