by Liam Vaughan
Between 2 January and 18 January, the trader had accumulated a long position of $70 billion, double the market capitalisation of the entire bank. As his colleagues left the trading floor each evening, Kerviel had stayed behind manically buying futures tied to the DAX and other indices, convinced that the worst of the crisis was over and that the markets would rebound. But his winning streak had come to an end. Kerviel’s wave of after-hours buying only ever propped DAX futures up for a few hours each night. Then, like some horrific Wall Street version of Groundhog Day, he awoke each morning to find gravity had kicked in and the market had sunk back in line with the rest of the world. As Kerviel made his confession, Société Générale’s management ordered one of his colleagues to close out his positions. By the time the employee was finished, the bank had lost $7.2 billion. News of the incident rocked global markets and helped push the DAX 12 per cent lower in two days, wiping hundreds of billions of dollars off the value of Germany’s biggest companies.
Reading about events at Société Générale, the traders at Futex quickly worked out that Kerviel had been the one behind the DAX’s strange manoeuvrings. It wasn’t the Chinese after all. One of Europe’s biggest banks had been brought to the brink by a lone trader with oversized ambitions and inadequate oversight. Later, Kerviel was sentenced to three years in jail and ordered to pay back the entire $7.2 billion he lost, the biggest fine ever levied on an individual. His desperate buying spree placed him among history’s most notorious rogue traders, a name uttered alongside the likes of Nick Leeson of Barings Bank and Kweku Adoboli at UBS. It also gave a young day trader from Hounslow the capital he needed to take his trading to new heights.
CHAPTER 5
RISE OF THE ROBOTS
The more money Nav made, the bigger the positions he took, as if he were climbing the levels of the ultimate computer game. In a matter of a few months, he’d gone from placing fifty- and hundred-lot orders to five hundred lots, more than anyone else at the firm. At that size every tick, or 0.25-point move, of the S&P 500 was worth $6,250. Becoming a bigger player had its advantages, but lately the markets had left Nav feeling unsettled. Beginning in around 2007, the ladder had become slippery and harder to read. Orders flashed then disappeared like phantoms, prices moved in unfamiliar ways and spoofing – the placing of bogus orders with no intention to actually buy or sell – had become rife. Nav’s opponents seemed to him to be so adept at predicting his next move that he became convinced they could ‘see through the ladder’ at who they were trading with. This sense of being persecuted coincided with the emergence of a new breed of participant engaged in something called high-frequency trading, or HFT.
HFT involves rapidly buying and selling assets using state-of-the-art technology to profit from extremely short-term price moves. The term is somewhat vague and ill-defined, but it’s used by some academics focused on futures to refer to a relatively small number of very active entities that trade large quantities of contracts without ever accumulating a sizable position, and that end each day flat. At the heart of HFT are algorithms, or ‘algos’ – sets of rules instructing computers to react to shifting market conditions with very little human intervention. These can be as simple as ‘if the price of x moves by y, buy z’, and as complicated as anything a roomful of nuclear physicists could come up with. Like human scalpers, HFT practitioners make decisions based on short-run shifts in supply and demand and the relationship between correlated markets, as opposed to any fundamental sense of the value of a share or commodity. But while the quickest-drawing day traders may be able to react to new information in a fifth of a second, the speed of HFT is measured in microseconds – millionths of a second.
The first HFT firms, such as Getco and Jump Trading, were started at the turn of the millennium by former traders from the Chicago pits who, instead of opening arcades and backing human traders as Paolo Rossi had done, hired teams of coders and maths PhDs to programme computers that could trade autonomously. Before long, they had descended on the markets like a plague of stinkbugs. In 2003, HFT firms barely registered in US futures markets, but by 2008 they were involved in one in five trades and by 2012 as many as 60 per cent. For a long time, HFT existed outside the mainstream consciousness. Its practitioners were small, privately owned, lightly regulated (most operated through licensed brokerage firms) and highly secretive. Even Wall Street was behind the curve. It was only during the financial crisis, when the leading HFT players started reporting profits in the nine figures, that people began to sit up and take notice. Articles appeared in the business press with headlines like ‘Rise of the Algos’ and ‘Robot Wars’. Stanford and MIT graduates bypassed Manhattan and Connecticut’s hedge fund country and packed their bags for the Windy City. The draw was obvious. In 2008, while much of the financial sector was in a battle for survival, Jump, a firm of a few dozen employees, made $316 million. The same year, Citadel, owned by one of America’s wealthiest and most secretive investors, Ken Griffin, pulled in about $1 billion in a unit devoted to HFT.
In spite of this flurry of attention, few people outside the close-knit industry really understood what these firms did or where the money was coming from. A bedrock principle of economic theory is that, the higher the returns of an investment, the riskier it will be, yet somehow the academics at these outfits, many of whom had no background in finance, had found a way to subvert that. By combining speed and statistical analysis with a unique understanding of the architecture underpinning electronic markets, they achieved the holy grail of investing, making large and consistent profits while taking very little risk. Hard data on HFT was hard to come by and firms were under no obligation to reveal their strategies, even to the regulators who, for a long time, seemed willing to take on trust that participants were acting responsibly. The only repository of real-time trading data was the exchanges themselves, private enterprises that straddled a curious line between seeking to attract the HFT firms’ business and being responsible for policing them in their role as ‘designated self-regulatory organisations’. Since HFT is predicated on vast numbers of transactions and exchanges collect a commission on every trade, their interests were closely aligned. Indeed, the explosion of HFT helped make the CME Group, which had only recently gone public, one of the most profitable companies in America. This symbiotic relationship was perhaps best illustrated by the revolving door between the top firms and the exchanges where they operated. Virtu Financial Inc.’s director, John ‘Jack’ Sandner, was the longest-serving chairman in the CME’s history. William Shepard, another longtime CME Group board member, is reported to own a sizable stake in Jump. The high-frequency trading industry and the CME Group were not just on the same side of the fence. They were the same people.
Executives at the likes of Jump, Citadel and Hudson River Trading pointed to the higher number of trades in general, and the shrinking gap between asking and buying prices, as evidence that their activities were improving the marketplace by making transacting cheaper and less volatile for everyone. But trading is a zero-sum game, and if HFT firms were winning, somebody had to be losing. In a research note suggesting institutional investors and pension funds were the ones getting stiffed, New York consultancy Pragma Trading wrote: ‘Given that HFTs are very short-term intermediaries between the directional traders who are actually trying to accumulate or unwind a position, it is hard to see how they can simultaneously be saving investors money and pulling billions out of the markets in trading profits.’ The small number of researchers who did manage to examine HFTs’ methods found evidence of practices including ‘momentum ignition’, a kamikaze attempt to incite abrupt market moves; ‘wash trading’, or trading with yourself to push the market around or pick up rebates; and ‘quote stuffing’, whereby entities flood the market with huge numbers of orders in a short space of time to cause delays. It was impossible to know just how widespread any of this was since the regulators, by their own admission, lacked the technology and expertise to monitor what was happening in their purview. There w
ere also questions about what such a fundamental shift in the makeup of the markets might mean for stability. ‘What happens if a major event causes turmoil in the market? Will these HFTs simply shut down their computers and walk away since their model has been corrupted?’ wrote Joe Saluzzi from New York brokerage firm Themis in a strikingly prescient 2009 blog post. ‘Where will all that LIQUIDITY that they claim they provide go when the market doesn’t suit them? A major vacuum will be formed in the market as multiple parties run for a much smaller than expected exit.’
For day traders like Nav, there were more pressing concerns. HAL 9000. The Matrix. Skynet. Dolores. The existential threat posed by robots has been rooted in the human psyche since the dawn of computers. For Nav and his peers, the fear wasn’t that they would upend the financial order altogether, but simply that these ultrafast, highly sophisticated machines would be able to scalp better and more efficiently than they could, crowding them out of the market. A poster on one of the trading forums drily captured the prevailing mood. ‘One day, years from now, a trading desk will typically consist of three things. A man, a dog and a computer. The computer’s job will be to trade. The man’s job will be to feed the dog. The dog’s job will be to bite the man if he goes anywhere near the computer.’
The exact strategies HFT firms employed were diverse and constantly changing, but at the heart of many of them were three elements: an ability to predict which way the market was about to move; the speed to capitalise on that move; and a novel way of minimising losses when they guessed wrong. The first element involved statistically analysing changes in the order book and elsewhere for information that indicated whether prices would rise or fall. Inputs might include the number and type of resting orders at different levels, how fast prices are moving around and the types of market participants active at any time. ‘Think of it as a giant data science project,’ explains one HFT owner. For years, Nav had used his superior pattern recognition and recall skills to read the ebbs and flows of the order book until it became second nature, but even the most gifted human scalper is no match for a computer at parsing large amounts of data.
When it came to speed, the leading HFT firms invested hundreds of millions of dollars in computers, cable and telecommunications equipment to ensure they could react first in what was often a winner-takes-all game. Exchanges charged tens of thousands of dollars a month to allow customers to place their servers next to the exchange’s to minimise any lag in receiving data. The result was, in the words of Eric Budish at the University of Chicago Booth School of Business, a ‘neverending socially-wasteful arms race for speed’. Only the top-tier firms could afford to keep up, meaning barriers to entry were high. And the exchanges gave their most valued customers sweetheart deals that slashed their cost of trading to a fraction of other participants’. Point-and-click traders had no chance. By the time a human being had seen, processed and reacted to a buy signal or news of an interest rate cut, the market had entirely absorbed the information and any value had evaporated. Tried and tested arbitrage strategies, whereby traders look for two closely correlated securities to temporarily fall out of kilter and bet they will come back in line, were fruitless in the face of machines that could identify and profit from anomalies tens of thousands of times quicker than they could.
The final element of HFTs’ success was an understanding of the plumbing underpinning electronic markets. The CME’s platform, Globex, is what’s known as a ‘First In First Out’, or FIFO, market. That means that whenever a trader places a limit order (that is, an order away from the current price), it joins the back of the queue, behind any other orders at that level. If the best bid in the market is currently $99.00 and you want to sell when it reaches $100.00, for instance, your order will be placed last in line behind any other traders currently looking to sell at $100.00.
HFT firms monitor these queues for opportunities to benefit from what is essentially a risk-free option. Consider an example: HFT firm AGGRO decides the market is statistically likely to fall and so places an order to sell ten e-minis at the current price (the ‘best ask’), again let’s say $100.00. As AGGRO’s ten-lot order makes its way to the front of the queue, new orders join at the same price; by the time AGGRO’s e-minis are purchased, or ‘hit’, another one thousand lots are waiting in the $100.00 line for a buyer. At this point, there are two possibilities. If the market falls, as AGGRO predicted, the firm can buy ten e-minis at a lower price, say $99.50, and walk away with a profit. If it starts to appear, however, based on fresh information entering the order book, that the price is actually going to rise, AGGRO can quickly turn around and buy ten e-minis from somebody further back in the $100.00 line, exiting the trade without taking a loss. By consummating trades only when they knew they could extricate themselves with little or no loss when they got it wrong, HFT firms largely eradicated losses. They also helped create a situation in which, by 2010, the overwhelming majority of all orders on the CME were cancelled before they were consummated.
With all this to-ing and fro-ing, it’s easy to understand how day traders like Nav came to believe they were being targeted. Every time they placed or cancelled an order, even if it was only a handful of contracts, the market moved. ‘I remember clearly the first time I noticed the HFTs,’ recalls one of Futex’s senior traders from that era. ‘It was the start of a new year, we logged on, and the order book just seemed subtly but discernibly different, like an update on your phone or something. I was on a bank of twelve desks and we all just looked at each other and said, “What is going on?” And from then on it became much harder.’ At arcades around the world, algos became like bogeymen, blamed for anything and everything that went wrong. If a trader took a position and the market moved against him, it wasn’t a bad trade, it was the ‘fucking algos picking me off’. Rumours abounded about illicit deals between the exchanges and the HFT giants, but the truth was that the robots didn’t need to know their opponents’ identities to thrive. Fast machines, cheap commissions and probability were enough.
The ascendancy of HFT squeezed many human scalpers out of the market. Some adapted by trading over longer time horizons, leaving positions running for hours or days rather than seconds. Others took to actively seeking out and trying to exploit algorithms, which quickly became ubiquitous among banks and asset managers as well as HFTs. In 2007, Svend Egil Larsen, a self-described algo hunter from Norway, noticed a flaw in the way an entity reacted to trades in certain stocks and set about taking advantage. He made a modest $50,000 but was later charged, along with a colleague, with market manipulation. The alleged victim in the case was a broker called Timber Hill, one of a raft of companies owned by the Hungarian-born electronic trading pioneer Thomas Peterffy, a man whose personal wealth is estimated by Forbes at $17.1 billion. Larsen was originally found guilty and given a suspended sentence, but the conviction was overturned on appeal. ‘We feel like Robin Hood, or David beating Goliath,’ he told the Financial Times.
Most day traders harboured a degree of resentment towards the HFTs, but for Nav, who had a fierce antiauthoritarian streak, it tapped into something deeper. How could he compete with a bunch of faceless billionaires who never lost? And how was that fair? The markets were supposed to be the ultimate meritocracy. It didn’t matter what you looked like behind your screens, or where your parents came from. If you made the right moves, you got the rewards. Except, Nav was increasingly coming to believe, that wasn’t true. Like so much else in life, the players destined to win were the ones with the most money and the right connections. In reality, Nav didn’t know who his opponents were, and it would later transpire that some of those he complained about the most were actually gifted human scalpers with limited technology just like him. But to his mind, they were all cut from the same cloth: privileged elites with better equipment hell-bent on trying to take him down.
With his maths skills, his aptitude for pattern recognition and his lateral way of thinking, Nav might have made a highly prized employee for an HFT firm. Instea
d, he made a decision that, as an already successful and wealthy trader, he didn’t have to make. On 4 June 2007, after three years of silence, That’s a Fugazi filed a new post. It was titled ‘S&P 500 Futures Corruption’, and it read:
For all those that trade the e-mini S&P 500 with a ladder (where you can see the bids and offers), you must have realised now how some market participants have an unfair advantage over the rest. I’m mainly referring to the two spoofers … who are there every day and seem to push around the market. Now, I’m not one to complain about spoofing I mean hey it happens in every market, but these two S&P spoofers CANNOT BE HIT. I’ve tried many times gentlemen and they simply can’t. Hence, they contravene the rules, as per CME themselves and should be eliminated from the market … I’ve spoken to CME about this and they simply refuse to accept that it is going on, even though you only need to watch the ladder at any time during the day to see that it is. It’s a clear example of letting the big guys get away wth blue murder at the expense of the small guys.