Courtesy of the author, we present to our readers the following excerpt from Dark Pools: High-Speed Traders, AI Bandits, and the Threat to the Global Financial System, by Scott Patterson, author of The Quants.
In the dark dawn hours of yet another frigid New York morning, Josh Levine dressed quickly and left his apartment in the shadow of the Twin Towers at Battery Park. He trudged through city blocks clogged with treacherous mounds of packed dirt-pocked snow and ice from the year’s record-breaking string of storms. It was February 16, 1996. A Friday. Earlier that week, IBM’s AI supercomputer Deep Blue defeated world chess champion Garry Kasparov for the first time. Bill Clinton was entering his second term in office. The country was in the midst of a powerful economic boom that would culminate in a massive tech-stock bubble, and Levine and Datek’s army of traders would be at the center of it.
The programmer walked east on Wall Street, then turned south on Broad, passing by the marble façade of the NYSE. At 50 Broad, he took the elevator to the sixth floor and walked past a few early risers, the wired, sunken-eyed day traders for Datek. Staring at their computer terminals, hugging their steaming coffees, waiting impatiently for the action to start at the opening bell, they nodded to Levine quietly as he passed.
Months earlier, Levine had decamped from the tiny cluttered room on the tenth floor of 50 Broad to a larger room on the sixth.
The room quickly slouched toward the en- tropic chaos Levine seemed to thrive in. He brought in an inflatable kiddie pool and populated it with turtles. He grew sea monkeys in a glass jar. An Israeli army bazooka leaned in a corner. The ubiquitous garbage piles, the tech ’zines such as PC World, stacks of computer books, pizza boxes, magazines, crunched Coke cans, crumpled computer printout paper, and candy wrappers rose to the ceiling like tropical plants, competing for space with the baker’s racks of computers, rows of computer terminals lined up on card tables, electric cords and creeping cables shooting out of the floor and through holes in the ceiling.
Levine shed his jacket, sat down before his monitor, and hit a but- ton. His computer hummed awake. It was time. The pieces were in place. Now he was about to bring to life the heretical idea that he’d been nurturing ever since he was a teenage runner in the 1980s.
Typing rapidly, he called up the program. Inputting a few instructions and taking a deep breath, he turned it on. Island was alive.
“Island is here!” Levine wrote on Watcher News. “You now have the ability to execute Island orders from the safety and comfort of your own Watcher.”
Wall Street would never be the same.
One of the earliest Island clients was a secretive, highly successful hedge fund based on Long Island called Renaissance Technologies. At first, Renaissance’s programmers—the firm was entirely run by mathematicians, scientists, and computer wonks—were dubious of Island. The reason: Datek. They were suspicious that the Datek bandits were secretly watching Island’s flow and front-running it.
But Island proved too big to ignore. One day in the late 1990s, several of Renaissance’s top executives, including a pair of AI experts who’d formerly worked at IBM, Peter Brown and Bob Mercer, paid a visit to 50 Broad. They were bemused by the giant lizard, the pool full of turtles, the trash. Renaissance’s headquarters in East Setauket, Long Island, were as pristine as an Ivy League campus. More than anything, Brown and Mercer were deeply impressed by Levine, whom they instantly realized was a programming genius with a profound understanding of the market’s plumbing.
Renaissance deployed cutting-edge AI programs to build its models and guide its trading. But Renaissance’s AI was an evolutionary leap beyond what any other firm had ever attempted, creating a strategy that would turn Renaissance into the most profitable trading machine in the world, with annual returns averaging north of 40 percent a year. Island’s high-speed platform was an ideal match for its strategies.
The Island-Renaissance fusion was a vision of the future in which high-speed AI-guided robots would operate on lightning-fast electronic pools, controlling the daily ebb and flow of the market. The AI Bots poured their valuable liquidity into Island, which, in turn, made it possible for the Bots to operate at high frequencies. They fed off one another, creating a virtuous cycle that would become un- stoppable. Little-known outfits such as Timber Hill, Tradebot, RGM, and Getco would soon start trading on Island, forming the emergent ganglia of a new space-age trading organism driven by machines. Tricked-out artificial intelligence systems designed to scope out hid- den pockets in the market where they could ply their trades powered many of these systems.
In the process, the very structure of how the U.S. stock market worked would shift to meet the endless needs of the Bots. The human middlemen, though they didn’t know it, were being phased out, doomed as dinosaurs. And the machines were breeding more machines in an endless cycle of innovation, as programmers pushed the boundaries of speed more ruthlessly than Olympic sprinters. Trading algorithms would mutate, grow, and evolve, feeding off one another like evolving species in a vast and growing digital pool.
In the 1990s, that future had been inconceivable to all but a few visionaries—such as Levine, who had a bird’s-eye view of the market as it evolved. But in an ironic twist, what the market would evolve into would be a market far darker than anything Levine could ever have imagined—a market dominated by secret trading algos, vast data centers, and dozens upon dozens of dark pools.
Because the market was so dark, conspiracy theories became rampant. Rumors spread that some firms were sending waves of orders to jam up exchange matching engines—the core computers that paired up buy and sell orders—hoping to profit on discrepancies in prices across multiple pools. Sophisticated tricks known as “layering,” “spoofing,” and “quote stuffing,” in which computers gamed the electronic pools with phantom orders, were said to be unbridled. Insiders talked of trading outfits based in countries such as China, India, and Russia that manipulated prices around the world through sophisticated computer games.
Neil Johnson, a University of Miami physicist who studies complex market patterns, warned in a February 2012 interview of a “global war between competing computer algorithms” that could cause a “big system wide collapse” in which the stock market shatters like broken glass. The market, he said, had evolved into a “lake full of different types of piranhas” devouring one another in a high-speed frenzy. In a working paper he co-authored—“Financial black swans driven by ultrafast machine ecology”—Johnson showed that the market in the past few years had undergone “an abrupt systemwide transition from a mixed human-machine phase to a new all-machine phase characterized by frequent black swan events with ultrafast durations.”
It was a brave new world that few people outside Wall Street knew existed. The Algo Wars were leaving a path of destruction in their wake. And what about the Securities and Exchange Commission? The nation’s top stock-market cops were clearly outgunned. In May 2011, SEC Chairman Mary Schapiro told Congress that “the Commission’s tools for collecting data and surveilling our markets are wholly inadequate to the task of overseeing the largest equity markets in the world.”
It was as if the FBI were admitting that it couldn’t track organized crime. Everyone, it seemed, was in the dark.