Machine Mania in the Marketplace: How Computers Came to Own the World

By David Haggith

With 60% of stocks now being traded by bots that fake each other out in order to create buying opportunities, stock exchanges have lost their connection to the reason markets are created in the first place. The exchanges no longer exist as places for people to buy and sell ownership in a corporation. They exist simply as the neural junctions of a conglomerated machine that plays tricks on itself, and your sole goal is no longer to invest, but to put money in the slot machine that is the quickest trickster.

Many of the people who think of themselves as investors see this pretend investing as being almost risk free now that computers and central banks are running the racket. They put their money in the machines, and machines follow the central banks’ lead, purring along at historically low levels of market volatility as the machines run their automated tasks. A minority of market experts see a market that is building cataclysmic risks as it accumulates fake pricing that has nothing to do with intrinsic value and as the component machines keep getting reprogrammed to do a better, faster job of faking out the other machines.

[Brad] Katsuyama, whose firm and company were made famous by Michael Lewis’s 2014 book, Flash Boys: A Wall Street Revolt, says computers running complex software conducting trades at lightening speeds [are] a “dangerous” threat to the stability of the market, juicing volumes and sparking so-called flash crashes, where assets swing rapidly in value in a matter of seconds. “I think the biggest risk in the market is that 50-, 60-plus percent of the volume is being executed by computer programs who have no idea what companies actually do. They’re just reacting to data. And I think it’s dangerous.” (MarketWatch)

Katsuyama, whose company is starting its own stock exchange to try to combat the machines, blames rare bouts of volatility (flash crashes) on the computer algorithms that now dominate market trading. For example, when Amazon lost $40 per share in four seconds on June 9th this year and then immediately recovered, Katsuyama says you can be certain that didn’t have anything to do with a change in Amazon’s intrinsic value nor with any fundamental economic changes in this world. Some algorithm somewhere jogged a price switch and caused other algos to sing in harmony, flash-crashing Amazon’s stock and triggering a general decline in high-tech stocks. A computer glitch? Or arcane trickery by which the brainiest bot at that particular nanosecond managed to trick all the other bots in order to create a dip and then buy the dip and make billions?

Yes, there are enemy bots that know the other bots, find and exploit their weaknesses to trick as many as they can into selling in one direction in order to buy the trade in the other direction. In fact, a virtual lexicon of slang is gaining popularity with terms like “wash trade,” “layering” and “spoofing” for the kinds of micro teases and diversions and tricks employed by enemy bots.

In a “wash trade,” a trader acts as both buyer and seller of a stock, to create the illusion of volume. “Layering” and “spoofing” are off-market orders designed to trick the rest of the market into thinking there are buyers or sellers of a stock waiting in the wings, in an attempt to nudge the stock price one way or the other. (Vanity Fair)

During the financial crisis of 2007-2009 that brought down the world, only 30% of assets were traded by computer-generated trades. At double that amount, the next time will be different. Old-fashioned traders who research companies to buy stocks based on perceived company value now account for about 10% of the US stock market. There is very little concern in today’s trading for economic and business fundamentals.

Corporations are now just toys to be played with by the machines.

The strange new world of undefinable, self-programming bots

The scary part is that no one seems to know what causes specific flash crashes in many cases. Even Katsumaya only guesses at what really happened during the Amazon flash crash because visibility in the world of trader bots (or traitor bots) is zero. By that, I mean that even the people who create these algorithms truly have no idea what the bots’ current programming is because the programming is designed to be perpetually self-modifying through some vague crocodilian artificial intelligence created at the cerebral cortex of their semi-simian brains.

While no one seems to know the cause of Amazon’s flash crash, the Nasdaq ended 1.8% lower that day. The bots know best, though they actually know nothing at all. They merely respond to targeted stimuli.

It is the same in bond trading as in stocks. When the market for US treasuries flash crashed in 2014, it triggered extensive studies that revealed high-frequency traders (HFTs) were the culprits. Of course, HFTs are computers that place zillions of trades based on zillions of micro calculations every day.

Bear in mind that these auto-traders were mostly designed by young people, fresh out of college who have never known a bear market during their adult lives. The machines that determine the “market value” of all the corporations in our world were programmed by people who are only familiar with the dynamics of an always-rising, central-bank-driven market. How well the machines work if the market ever finds a way to slip into reverse, no one knows. The algos have never been tested in a true bear market as to how they might team up to accelerate the market’s decline. Since they were taken out of the box, they have been reprogramming themselves entirely based on bull-market dynamics. How they work running downhill is purely theoretical and unknown even then because of their self-programming nature.

But it will be fine. Trust the machines and their child creators who have little depth in the real-world markets.

Also unknown in this realm are the hackers lurking beneath these murky waters — be they anarchists or Korean agents or teenage savants seeking instant wealth — who might exploit the weaknesses and strengths of the machines in order seize ownership of the corporate world bit by byte or all in one colossal dump.

Kill switches are the only safety between people and economic ruin

The creators have lost control of their creation, except for the ultimate solution of simply pulling their plugs on any given day and/or closing the markets; but there are safeties built in. The robo-traders operate on set parameters. When things happen outside those set parameters, they either don’t exist in the bots’ perceptions — at least, not until they move things that do exist within the parameters — or the bots flee the market, whirring away on standby, until they understand what is happening. (“This does not compute.”) Suddenly volume and trades hit the floor. 90% or more of trading can instantly vanish.

When that happens, prices go wonky as the remaining few algos go wild. Their resulting erratic trading spikes volumes and prices all over the place…. The flash crash in May 2010 [a one-thousand-point plunge] gave us an indication, but the mini flash crashes we see almost daily in various other markets — ranging from the tiny to the US Treasury market — tell us that it’s entirely possible that someday all the worlds computer algos might suddenly stop operating because an event occurs that is out of their programmed operating state….  (Peak Prosperity)

Not responding to forces that don’t compute under the machines’ bull market bias could be design safety, or it could just be blindness to outside forces. On the days of major flash crashes when no one at the exchanges understands what is happening, the solution has been to shut down the market or start pulling plugs on the bots. Even though the exchanges may have automated stops, the trader bots may be faster than the stops or be designed in some nefarious way to circumvent the stop. That means the losses from mere bugs could potentially be astronomical, with errors replicated millions of times per second and repeated by unknown bots trading from all over the world until someone physically throws the main breaker to shut down the entire exchange.

When a crash does happen, and the final solution is employed (taking the market offline), the market cannot go back online until parameters somehow level back out to where reconnecting the bots for trading doesn’t instantly jog the crash to an even deeper level.

If computers lose control and take the market deeply down, the market may reopen in a price vacuum — a position where the few remaining human investors have no clue what the real value of any stock should be because true price discovery has been dead for years and where the machines know only the closing point and see nowhere to go but down. So, a reset may not be easy.

What happens if the machines have to be kept offline until they can be reprogrammed and tested offline to handle a bear market without creating further disaster? Does the market still reopen to the minority of human traders doing things the old-fashioned way? How can it? The normal price-earnings ratio for stocks has been around 18x. With Amazon now at 188x, what’s the right price if the market does a major reset based on fundamental values, instead of mechanized speculation?  90% lower? That’s where the old-fashioned traders might take it. The truth now is that, if fundamentals ever return to the investment equations, the result would be catastrophic revaluation.

So, the same machines jump in all at once and to decide the opening price after a devastating plunge, and major corporations become a victim of whatever last price the machines set on the casino tumblers, having nothing to do with their business value. At this point, the market is almost entirely a speculators’ market, and the speculators are almost entirely machines.

When a sell-off does occur, most machines will be doing the same trade in the same direction because they have all been buying the same stocks for a long time (hence Amazon’s 188x price-earnings ratio). The risk whenever too much capital is being committed in too few places at the same time is that the exits are few and narrow. When every machine wants to sell the same handful of stocks and no one wants to buy those stocks because the machines are still dumping, the resulting price vacuum will form a downward vortex that should be … well, fascinating to behold.

Andrew Lapthorne from Societe Generale describes the process this way:

At some point, that reversion process will take hold. It is then investor ‘psychology’ will collide with ‘margin debt’ and ETF liquidity. It will be the equivalent of striking a match, lighting a stick of dynamite and throwing it into a tanker full of gasoline. When the ‘herding’ into ETF’s begins to reverse, it will not be a slow and methodical process but rather a stampede with little regard to price, valuation or fundamental measures.

Importantly, as prices decline it will trigger margin calls which will induce more indiscriminate selling. The forced redemption cycle will cause catastrophic spreads between the current bid and ask pricing for ETF’s. As investors are forced to dump positions to meet margin calls, the lack of buyers will form a vacuum causing rapid price declines which leave investors helpless on the sidelines watching years of capital appreciation vanish in moments. (TalkMarkets)

Remember, the investors here are now almost all computers, so this stampede of price changes can happen at the speed of light. When it’s time to head for the door this time, you have to be faster than nanosecond exchanges of the computers that are all trying to get the jump on each other. Good luck with that!

Why bots don’t know how to navigate crashing markets without crashing them worse

The simple truth is that bots learn as they go, and they all learned by trial and error in normal markets where the norm has been an upward trend line. Because market crashes are rare, it’s also hard to find enough data to train a bot how to run for the betterment of humanity or just the betterment of its owners during a crash situation. Even if you gather enough data about past crashes, you have to wait for the next crash to test your programming in a real situation. In the meantime, everyone else has been attempting to program their bots for the next crash — some to avert a crash, others to exploit it — so by the time a real-market crash comes along for testing your own modifications, all the devices that tumble the data are behaving in a different manner anyway.

Sudden spikes in volume and price cause spikes in volatility, and a market where volatility now slumbers at the bottom of the swamp is especially prone to exaggerated disruption and catastrophic failure. The reason for that is that a large part of the market is trading volatility. Hundreds of billions of dollars in assets are linked to volatility. A spike in volatility, even to historic norms, will cause significant volatility selling. Such selling increases the volatility, triggering more selling, and bots can amplify the speed of all of that exponentially.

Moreover, the longer the market has run on low volatility (due in part to bots running the market in highly programmed ways), the more leveraged the market becomes as people take on more debt in their bets when they perceive risk is stabilized at a low level and that trades are running consistently in one direction. Low volatility causes people to presume lower risk. People simply go further out on a limb, taking out bigger loans to gamble. Obviously, investments made on loan make for an excruciating situation if the market falls.

As a result of historic extended low volatility, markets right now are sleep-walking past risk. I would suspect they are almost numb to pain, believing central banks will save them from anything because central banks appear to be steering the market toward constant success (see next section below).

Leverage is its own accelerant to any market conflagration. Leverage plus thousands of super-computer bots, could be a market atom bomb.

This environment has amassed phenomenal risks. These shifts, as we have seen with the tech stocks, can occur without prior notice, without obvious trigger. They occur because an algo sets it off and other algos follow since they react to each other, and the whole machinery can suddenly go into reverse and get stuck in it. (Wolf Street)

Central Banksterbots

The markets are also low in volatility because central bankers know who to tease the right bots with the right price points to herd the other bots in the right direction.

A long-running discussion between Dave Fairtex, myself and others, concerns the idea of whether or not markets as big as the ones just mentioned can be manipulated by government/central banking forces to stop, limit, or even reverse a price decline.

My view has always been “yes”, because it should be child’s play to fool the algos into going this way instead of that way by simply injecting a relatively small amount of capital at the right place and time.

I would love to know, for example, why central banks have an incentive program at the CME — where the exact sorts of highly leveraged, electronically traded products that would be best suited for market manipulation — are traded.

By virtue of its existence, we know that central banks are highly active traders on the CME platforms.  Otherwise an incentive program offering steep volume-based trading discounts would not exist.

Not one single central bank (yet) reports anywhere in their financial disclosures of being the proud owners of any of the accounts traded on the CME. So the details of the situation remain a mystery.

But dependably, every single market decline that began over the past several years has been reversed — usually in the dead of night, and in the futures market — by mysterious injections of capital that then get the HFT algos to follow the trend.  So inquiring minds would like to know. (Peak Prosperity)

While the central banks know how to tease everyone else’s bots into a preferred direction of trade, their member banks have their own banksterbots, also trying to direct trade to the banks’ best interest.

JPMorgan has decided to eliminate carbon-based traders entirely. The Financial Times reports that JPM will conduct trades across all exchanges with a new super computer under new artificial-intelligence programming that has proven far more efficient than a host of human traders. (I wonder if they could also program the computer to become their new CEO, saving a huge expense on CEO salary, benefits, and bonuses?) Like all trader-bots, JPM’s behemoth will execute trades based on billions of transactions that it has learned from. One of its key advantages over competitors is that it has learned how to offload billions in assets without changing market prices.

JPM believes their new technology places them almost two years ahead of all their competitors technologically. And technology seems to be the only trading advantage companies now have in an investment environment almost devoid of human beings. More and more banks and brokerages are investing in faster wires, rather than better-schooled employees. Speed is parsed so finely that, even for transactions moving at the speed of light, traders using servers located closer to their markets can outmaneuver everyone else. Winning in this market is entirely about transmission speed.

The banks are not just cutting costs by removing people; they are pressed to focus money entirely on technology. So, even as the central-intelligence of the banking system (the Fed) has claimed that its financial engineering is intended to improve wages, it is hard to see how that is going to happen when the institutions closest to the Fed’s program are all downsizing humans in order to upsize computers.

UBS is another bank that is switching from human traders to super-computers, saving forty-five human minutes for client trades the super-computer can process in microseconds. UBS is particularly using its computer(s) to help clients trade volatility. No chance, I suppose that the artificial intelligence in the UBS-bot will find a way to spike the VIX to UBS client advantage and the rest of the world’s loss.

Heck! Forget the clients even. Soon they will be a useless as human traders. Being a stock or bond trader used to be a profitable career, but now it is the center of downsizing.

You’ll be comforted in knowing that JPM sees no risk-management issues with its new artificially intelligent super-computer.

The machine is restricted in its trading behaviour, as it learns under, and operates within, our general electronic trading risk framework, which is overseen by internal control groups and validated by regulators.

What could possible go wrong? Sounds as failsafe as Fukushima to me. Anyone reading here knows that anytime I hear humans talking about something they have created being “failsafe,” I become certain cataclysmic destruction is whetting its fangs right around the corner. No one could ever figure out how to hack that computer to get around the sleeping or bought-off regulators and apply its artificial intelligence toward ill-gotten gains, right?

Robotellers also rule

Banskterbots are not limited to use as traders.

Sweden, one of the pioneers of cashless economies, is switching to running its cashless economy with humanless devices.

What could go wrong?

Sweden has Aida, a sweet banksterbot that takes care of client transactions. Aida is a rapid study and always courteous, 24/7. She is your virtual customer-service representative extraordinaire at SEB AB, one of Sweden’s major banks. Their competitor, Nordea Bank AB, has Nova, and another competitor, Swedbank AB, has Nina. Computers are so much nicer when they have human names.

All three wonderbots will talk to you in a sexy female voice because female voices have been found to be more calming and appealing to customers. Aida even looks Swedish. What could be sexier? (I mean who doesn’t love Siri’s voice? She is filled with the most smart-alec answers to your love questions, even though she gives a fairly high number of dumb answers to important questions compared to her competitors. She is quite witty when I ask her things like when she went on her last date; but, if I want to find a shopping mall close to Toledo, Siri will give me directions to the nearest mall in Washington, D.C. She’s sporty that way, always trying to send me on wild goose chases around the country just for the fun of it, even when I tell her to use my current location.)

Aida may be an anachronism in her own time because it is not clear that banks even need customers anymore. If they don’t need customers, they don’t need customer service. Customer satisfaction with Swedish banks has dropped to a twenty-year low, but that is where these megabots are intended to help:

“Basically all banks are closing branches,” Mattias Fras, head of Robotics, Strategy and Innovation at Nordea, said in a phone interview. “This is a way to return to full service again.” (Bloomberg)

Ah, yes, full service by a machine as smart at answering your questions as Siri. What could be better for customers service than closing down convenient nearby branches full of human beings? Who wouldn’t prefer getting trapped in a robotic phone menu? Of course, it is about “full service!” (Gotta love the never-ending lies about what is good for YOU.) This couldn’t possibly have anything to do with making more money for stockholders while dumbing down your expectations for service. With banks making most of their money by training their algos to play with central-bank free money, customers are almost useless appendages in the whole process. So, of course, you’re going to be dealt with by automation, and the banks will try to convince you it is what you really want.

Given the penchant in Sweden for sexy female names that start with “N,” the next banksterbot will probably be named “Nora,” and she’ll be infected with a virus that is asexually communicable to all the other banksterbot supercomputers. If you want comfort regarding your bank’s increasing automated handling of you, just remember the ’60s adage (if you’ve been around that long) that computers don’t make mistakes, humans do; and we’re getting rid of them.

What could go wrong? I know I always love banging my iPhone against my desk as I tell Siri how dumb her latest answer was while thanking her for sounding so sexy in its deliverance.

It’s a robot’s world! Even computer wizard Elon Musk constantly warns of the dangers of AI.

And its not just banks and bonds and stocks. It’s also commodities.

If it’s not banksterbots, its gangsterbots

Here’s one man’s evening experience in the seek-and-destroy dens of the traderbots:

Check out the price action in natural gas futures last Thursday evening, the same night of the dollar flash crash.   We are sick and tired of this blatant market manipulation and a lot poorer from it.

We put on a short position in nattie Thursday night before driving back from Sacramento to the Bay Area.   We checked the market at dinner and see its down about 1 percent, we feel happy and give high fives.   Then we look at the position and we have none!

It was taken out (buy stop) as  a Seek and Destroy Bot came in around around 9 pm, guns the market to the upside to take out all the buy stops of the short sellers, then turns around and guns it to the downside to destroy all sell stops of longs.  Finally,  moves the market back to where it was before the all the nonsense began.  Not a bad day’s work for the robot.

…Nattie’s total move in those few minutes last Thursday evening was almost 4 percent…. This is total B.S.! This is not the first and only time we have lost money due to market manipulation.  (Global Macro Monitor)

Sighs. What is a trader to do when his own bots appear to work against him? A single tweet can cause any market to go into nano-free-fall.

The SEC recently charge UBS with high-frequency trades that tricked its own customers in its own dark pool (private internal stock market) by creating secret orders beneath the scummy surface in order to exploit its own clients. Barclays is another bank that has been charged with lying to investors about secret high-frequency traders in its own pool, which made it easier for the HFTs to trade against other investors.

Can you believe that all thirteen US stock exchanges are being sued in a class-action for cheating ordinary investors by selling privileged access to the HFTs that compete against the ordinary investor? As a result of the investigations, Bank of America, Citigroup and Wells Fargo have shut down their HFTs and dark investment pools. Who would have though sharks lurked deep within those dark waters?

Michale Lewis, who wrote Flash Boys, deduced in Vanity Fair,

It would have been difficult to find anyone, circa 2009, able to give you an honest account of the inner workings of the American stock market—by then fully automated, spectacularly fragmented, and complicated beyond belief by possibly well-intentioned regulators and less well-intentioned insiders. That the American stock market had become a mystery struck me as interesting. How does that happen? And who benefits?

By the time I met my characters they’d already spent several years trying to answer those questions. In the end they figured out that the complexity, though it may have arisen innocently enough, served the interest of financial intermediaries rather than the investors and corporations the market is meant to serve. It had enabled a massive amount of predatory trading and had institutionalized a systemic and totally unnecessary unfairness in the market and, in the bargain, rendered it less stable and more prone to flash crashes and outages and other unhappy events.

Subsequent to the publication of Flash Boys, numerous investigations by various agencies have begun.

The Financial Industry Regulatory Authority announced it had opened 170 cases into “abusive algorithms.”

What could possibly go wrong  … except that which has already gone spectacularly wrong before? It all happens by our allowance and our creation. We learn nothing from each flash crash but continue to allow the bots to run the show, recent investigations not withstanding, since previous investigations have not resulted in much jail time and certainly not in any important revisions to the robomarket. So, we will get another chance to learn it again soon. History is rife with repeated educational opportunities.

You can read more from David Haggith at his site The Great Recession Blog.