第 1 章
Wall Street's Hidden Battlefield: Where Milliseconds Mean Millions
In a world where the average investor still pictures the stock market as a bustling trading floor filled with shouting men in colored jackets, Michael Lewis's "Flash Boys" reveals a startling truth: that image is dangerously outdated. Today's markets operate inside black boxes in nondescript buildings across New Jersey, where milliseconds-even microseconds-determine who wins and who loses. The book became an instant cultural phenomenon upon its 2014 release, sparking congressional hearings, SEC investigations, and heated debates across financial media. It's been praised by Warren Buffett, who called it "a great story of how our markets evolved," and condemned by high-frequency trading firms who saw their secretive practices exposed. What makes this book so compelling isn't just its revelations about market structure, but the human story of unlikely heroes who discovered that the world's most important market had been rigged-and decided to do something about it.
第 2 章
The Invisible Line That Changed Everything
By summer 2009, an unprecedented and secretive construction project was carving its way through America's heartland. Two thousand workers, divided into 205 specialized crews, were engaged in what appeared to be an impossible task: creating the most direct fiber-optic line ever constructed between Chicago and New York. These teams worked around the clock, employing every conceivable construction technique - from traditional trenching and drilling to advanced mountain-boring technology and sophisticated underwater installation methods for river crossings.
This ambitious venture was Spread Networks, conceived by Dan Spivey, a former trader who had made a startling discovery while studying telecommunications maps. He noticed that existing fiber routes, constrained by following railroad tracks, created significant inefficiencies by meandering through the Allegheny Mountains. Through meticulous planning and innovative routing, Spivey determined that by utilizing secondary roads, remote dirt paths, and carefully negotiated private property easements, he could eliminate over 100 miles from the traditional route. What began as his casual observation - "I'd like to see how much faster someone would be" - evolved into a $300 million infrastructure project that would revolutionize high-frequency trading.
The project's secrecy was paramount and orchestrated with military precision. Multiple shell companies were established to obscure the true purpose and scope of the construction. Even Steve Williams, the veteran project supervisor hired to oversee the fiber installation, was initially given information about only a small segment of the route. The truth about the project's full scale - stretching all the way to New Jersey - was revealed to him gradually. The construction pace varied dramatically: crews blazed through the flatlands of Indiana and Ohio at an impressive 2-3 miles per day, but progress crawled to a mere few hundred feet daily when confronting Pennsylvania's notoriously dense "blue rock" limestone formations. Spivey's obsession with speed manifested in countless micro-optimizations, including his memorable complaint about road crossings: "Steve, you're costing me a hundred nanoseconds. Can you at least cross it diagonally?"
The project's unveiling to Wall Street in March 2010 sent shockwaves through the financial industry. High-frequency trading firms immediately recognized the competitive implications - access to this line wasn't just an advantage, it was a necessity for survival in the microsecond-driven trading world. The pricing structure was unprecedented: $300,000 monthly fees plus substantial up-front costs, approximately ten times higher than standard telecom rates. The first 200 firms would need to commit $14 million each for five-year contracts, potentially generating $2.8 billion in revenue.
The response from major financial institutions revealed the complex dynamics of Wall Street's technology arms race. Citigroup made the peculiar request to reroute the line directly to Manhattan, while Credit Suisse balked at restrictions preventing client access sharing. Morgan Stanley sought "plausible deniability" while accepting the strict terms, and Goldman Sachs readily embraced the conditions. The project raised a profound question that would fundamentally challenge Wall Street's operating principles: How could a mere millisecond advantage in data transmission command such astronomical prices? The answer would expose the hidden transformation of financial markets into a high-speed technological battleground.
第 3 章
When Markets Vanish Before Your Eyes
Brad Katsuyama had built a comfortable life at the Royal Bank of Canada. The son of Japanese immigrants to Canada, he'd risen to run RBC's equity trading department in New York, earning nearly $2 million annually in a bank known for its stability and ethics. Unlike many Wall Street institutions, RBC had resisted subprime temptations and weathered the 2008 financial crisis relatively unscathed.
But by 2007, Brad noticed something strange happening in the markets. When his screens displayed 10,000 shares of Intel at $22 and he pushed the button to buy, the offers vanished. This phenomenon became unmistakable when he tried to sell 5 million shares of Solectron after a buyout announcement. Though the screens showed he could sell a million shares at $3.70, when he attempted to do so, the price collapsed as if someone knew his intentions before he'd fully expressed them.
When Brad called tech support, they initially dismissed his concerns as "user error." Eventually, they sent developers who watched as he demonstrated the problem: he'd count aloud to five with his finger over the Enter button, and nothing would happen until the moment he pressed it-when all offerings would instantly disappear and the price would jump. "You see," he told them, "I'm the event. I am the news."
The stock market itself had fundamentally changed. When Brad arrived in New York in 2002, 85% of trading happened on the NYSE with human processing. By 2008, there were thirteen different public exchanges, mostly in New Jersey, all operating as computer servers with matching engines. The exchanges had also implemented a complex "maker-taker" fee model that few understood. Traders who "crossed the spread" were "takers" and typically paid fees, while those who placed resting orders were "makers" and often received payments.
Brad couldn't understand why exchanges would pay anyone to be a taker or why anyone would pay to make a market. When asking around, he learned about firms like Getco that controlled 10% of the U.S. market-yet Brad had never heard of them despite running a Wall Street trading desk. The lightbulb moment came: the markets were rigged, and the answer lay beneath the surface of the technology.
第 4 章
Uncovering the High-Speed Predators
Brad assembled a team to investigate the U.S. stock market, deliberately choosing technical experts rather than traditional traders. His unconventional recruitment strategy brought together Billy Zhao, a meticulous former Deutsche Bank programmer known for his deep understanding of exchange architecture; John Schwall, a Bank of America veteran with extensive back-office operations experience; Dan Aisen, a brilliant Stanford graduate with expertise in mathematical modeling; and Allen Zhang, an eccentric Chinese programmer whose unique problem-solving approaches would prove invaluable. Most crucial was Rob Park, a fellow Canadian whose reputation at RBC stemmed from creating sophisticated trading algorithms that could replicate human decision-making with remarkable accuracy.
The team initiated a methodical investigation into the mysterious phenomenon of vanishing orders. They spent weeks conducting controlled experiments, systematically testing different order types and routing combinations across multiple exchanges. After numerous conversations with exchange representatives and thousands of test trades, they identified a striking pattern: the success rate of order execution decreased proportionally as they distributed orders across more exchanges. BATS exchange stood out as an anomaly, consistently achieving 100% fill rates regardless of order size or market conditions.
Rob Park's eureka moment came during his morning shower - a revelation about the physics of electronic trading. He realized the critical factor wasn't just speed, but the varying distances between exchanges. Orders reached BATS in 2 milliseconds, while taking 4 milliseconds to reach Carteret, creating microscopic windows of opportunity. Allen Zhang developed sophisticated software that precisely calculated and implemented compensating delays, ensuring orders would arrive simultaneously at all destinations. Their initial test run produced unprecedented results - every screen displayed green confirmations, indicating complete order fulfillment across all venues.
The implications were profound and disturbing: sophisticated traders were exploiting these microsecond differences to effectively front-run orders between exchanges. Brad's response was immediate and ethically unequivocal. Despite the potential for enormous profits by exploiting this knowledge, he instantly decided to expose the practice, declaring, "We have to go on an educational campaign."
The team developed Thor, an innovative tool that incorporated precisely calibrated delays into stock exchange orders to ensure simultaneous arrival. The name quickly evolved into trading floor vernacular, with traders shouting "Thor it!" whenever they needed to protect large orders from predatory algorithms. Thor also provided unprecedented transparency into the cost of high-frequency trading practices. In a landmark test, they executed a 10-million-share Citigroup order, documenting savings of $29,000 by preventing front-running. While this represented less than 0.1% of the trade value, when extrapolated to the entire market's daily volume of $225 billion, it revealed a staggering $160 million in daily predatory trading profits.
"It was so insidious because you couldn't see it," Brad explained to investors and regulators. "People are getting screwed because they can't imagine a microsecond." This invisible tax on trading had remained hidden precisely because its timeframe existed beyond human perception, operating in the realm of microseconds where only machines could detect and exploit these opportunities.
第 5 章
The Telecom Expert Who Connected the Dots
Ronan Ryan didn't fit the Wall Street trader mold-pale, narrow-shouldered, and lacking the typical trader's confidence and swagger. Born in Dublin and relocated to America at sixteen, he'd struggled to break into finance and instead built a career in telecommunications infrastructure.
Despite no technical background, Ronan became fascinated with telecom infrastructure-how copper circuits compared to glass fiber, which buildings could support heavy equipment, and how information traveled through complex, zigzagging networks rather than straight lines. "When you make a call to New York from Florida, you have no idea how many pieces of equipment you have to go through for that call to happen."
By 2005, Ronan was helping high-frequency trading firms solve their "latency" problems-the milliseconds it took to complete trades between different locations. This triggered an arms race for speed. Traders obsessed over microseconds, demanding shorter cable routes, faster switches, and premium positioning inside data centers. When exchanges realized traders would pay for proximity, they began selling "co-location" services, placing trading computers directly inside exchange buildings.
The competition grew absurd-firms wrapped servers in wire gauze to hide improvements, left toy store logos on premium cages to disguise their position, and constantly upgraded equipment for minuscule advantages. "Never before in human history have people gone to so much trouble and spent so much money to gain so little speed."
Despite earning hundreds of thousands annually building speed systems, Ronan remained troubled by his limited understanding of why speed mattered so desperately. "I felt like the getaway driver," he explained. "Each time, it was like, 'Drive faster! Drive faster!' Then it was like, 'Get rid of the airbags!' Then it was, 'Get rid of the fucking seats!' Towards the end I'm like, 'Excuse me, sirs, but what are you doing in the bank?'"
When Brad Katsuyama offered Ronan a job with the title "Head of High-Frequency Trading Strategies"-despite Ronan knowing nothing about HFT strategies-he accepted immediately, excited to finally work in finance rather than technology. Though it paid only $125,000, a third of his current salary, Ronan saw it as an opportunity to understand what was really happening in the markets.
第 6 章
The System Behind the System
John Schwall, the son of a Staten Island firefighter who had risen to Head of New Products at Banc of America Securities, brought a dogged investigative spirit to Brad's team. His blue-collar background and twenty years of Wall Street experience gave him a unique perspective on market inequities. The financial crisis and Bank of America's acquisition of Merrill Lynch had profoundly upended his worldview when he witnessed Merrill executives who had created toxic mortgage bonds award themselves massive bonuses-some exceeding $10 million-while loyal Bank employees, including many of his colleagues, lost their jobs and retirement savings. "Wall Street is corrupt," Schwall concluded after watching the aftermath. "There is no corporate loyalty to employees."
Schwall's greatest obsession became understanding how the U.S. stock market had been rigged. Working late into the night, fueled by mounting frustration, he declared, "It really just pissed me off that people set out this way to make money from everyone else's retirement account." After spending countless hours researching "front-running" and "Wall Street," he discovered Regulation National Market System (Reg NMS), a seemingly innocuous rule passed by the SEC in 2005 and implemented in 2007. This discovery would prove pivotal in understanding market manipulation.
The regulation fundamentally changed how trades were executed by requiring brokers to find the best market prices for investors, replacing the previous loose standard of "best execution" with the legal requirement of "best price." This system relied on the National Best Bid and Offer (NBBO), calculated by the Securities Information Processor (SIP)-essentially the market's central nervous system. However, Reg NMS contained a critical loophole that would prove devastating: it failed to specify the speed at which the SIP needed to operate. This oversight allowed high-frequency traders to set up computers inside exchanges and build their own faster versions of the SIP, gaining a crucial 25-millisecond advantage over ordinary investors-an eternity in modern trading.
The gap between the public SIP and private feeds created immensely profitable opportunities for HFT firms. UC Berkeley researchers conducted a groundbreaking study that found Apple stock prices differed between the SIP and faster channels 55,000 times in a single day-effectively creating 55,000 opportunities for high-frequency traders to exploit slower investors. Using their speed advantage, they could buy shares at outdated prices and immediately sell them at new, higher prices, essentially guaranteeing profits with virtually no risk.
Through meticulous LinkedIn investigations, Schwall identified approximately twenty-five "kingpins" who truly understood the high-frequency trading ecosystem. He noticed a dramatic shift in the industry's composition, from traditional Wall Street traders to technical specialists-predominantly Chinese, French, Russian and Indian engineers with advanced degrees in physics, computer science, and mathematics. Credit Suisse provided perhaps the most blatant example of this duplicity: while publicly claiming their dark pool Crossfinder had nothing to do with high-frequency trading, their employees' LinkedIn profiles openly boasted about "building high-frequency trading platforms" and "implementing high-frequency trading strategy." This disconnect between public statements and private actions became a recurring theme in Schwall's investigation, revealing how deeply embedded these practices had become in the financial system.
第 7 章
The Russian Programmer Caught in the Middle
Sergey Aleynikov left Russia in 1990, not out of eagerness for America but because the Soviet system wouldn't allow him to study computer science due to his Jewish heritage. Starting with an $8.75/hour programming job at a New Jersey medical center, he worked his way through Rutgers University and eventually to IDT, a telecom company where he became the star technologist designing systems to route millions of phone calls.
In 2007, despite initial reluctance, Sergey interviewed with Goldman Sachs after a persistent headhunter convinced him to consider Wall Street's lucrative opportunities. Goldman's intense interview process surprised him with its competitive energy. He noticed that over half the programmers at Goldman were Russians, who had a reputation as Wall Street's best programmers-a skill he attributed to having learned to code efficiently with limited computer time in Russia.
Sergey joined Goldman during a pivotal moment, as its bond trading department was fueling a global financial crisis while its equities department adapted to radical market changes. The American stock market was fragmenting into thirteen public exchanges and eventually over forty dark pools, all trading the same stocks. This fragmentation generated massive trading volume from high-frequency trading firms.
Goldman's robots were slow, preventing them from winning the high-frequency trading races where winner takes all. Their system wasn't coherent but an amalgamation built on fifteen-year-old software, with approximately 60 million lines of code patched together like "a giant rubber-band ball." Serge was hired to solve three key problems: creating a faster ticker plant to process exchange data, improving the trading algorithms, and enhancing order entry systems.
Sergey discovered a gulf between his engineering mindset and Goldman's business culture. No one at Goldman had a global view of their software-there was shockingly little documentation, and they couldn't explain how components interacted. The environment didn't encourage good programming, which requires collaboration. "Everyone lived for the year-end number," with individual rather than team bonuses creating a fiercely competitive atmosphere.
When headhunter Misha Malyshev offered Sergey over a million dollars to build a trading platform from scratch for his hedge fund, he accepted. During his final six weeks, he mailed himself source code he'd been working on, containing both open source and Goldman proprietary code, hoping to disentangle them later. Shortly after starting his new job, three FBI agents arrested him for stealing computer code owned by Goldman Sachs. The prosecutor denied him bail, claiming he possessed code that could "manipulate markets in unfair ways." He was sentenced to eight years in federal prison without parole.
第 8 章
Building a Fair Exchange from Scratch
By late 2011, Brad had hit a ceiling with Thor. Despite RBC becoming America's top-rated stockbroker, they remained only the ninth best-paid. Investors loved Thor but felt obligated to allocate most trades to major banks like Goldman Sachs and Morgan Stanley to maintain relationships. As a Citadel trader told Ronan: "I know what you're doing. It's genius. And there's nothing we can do about it. But you are only two percent of the market."
Brad decided to take a radical step: create his own stock exchange. After selling his bosses at RBC on the concept, he approached major money managers and hedge funds who loved the idea but insisted such an exchange needed independence from Wall Street banks to be credible. Brad realized he'd need to quit RBC, find funding, and convince highly-paid professionals to work for reduced salaries.
On January 3, 2012, Brad arrived at work at 6:30 in the morning, resigned, and emailed his team. He left with nothing-no papers, no code-aware that Wall Street was using fear and legal threats to control technologists who understood the value of customer orders in dark pools. Though big investors controlling roughly one-third of the U.S. stock market supported Brad's mission, many questioned his motives: "Why are you playing Robin Hood?" For six months, Brad faked greed he didn't feel to put money people at ease, refusing Wall Street banks who wanted to invest to maintain independence and credibility. Eventually, he raised $9.4 million from nine money managers by December, with another $15 million six months later, plus his own life savings.
The new exchange, named Investors Exchange (IEX), needed people who could ensure it couldn't be exploited by predatory traders. The team analyzed hundreds of order types used by existing exchanges, discovering they were primarily designed to benefit high-frequency traders at investors' expense. They identified three main predatory HFT strategies: "electronic front-running" (seeing orders in one place and racing ahead elsewhere), "rebate arbitrage" (gaming exchange kickbacks without providing actual liquidity), and "slow market arbitrage" (exploiting price changes across exchanges before they could update).
To prevent predatory trading, IEX needed to be extremely fast while simultaneously slowing down everyone else. They created private direct connections to all exchanges using "the fastest subterranean routes." The 350-microsecond delay functioned like a head start, ensuring IEX could see and react to market changes before even the fastest traders could exploit them. To create this delay, they coiled thirty-eight miles of fiber into a shoebox-sized compartment, effectively banishing high-frequency traders to "West Babylon, New York" from their perspective, neutralizing their speed advantage.
第 9 章
The Moment of Truth
When IEX opened on October 25, 2013, the team's expectations were modest. Their median estimate was 159,500 shares the first day and 2.5 million the first week. They needed to reach about 50 million shares daily to cover costs. Their first day brought 568,524 shares, mostly from regional brokerages and Wall Street firms without dark pools. By mid-December, they were trading roughly 50 million shares weekly. Goldman had connected but was sending only tiny test orders that rested for milliseconds before disappearing.
Everything changed on December 19 at precisely 3:09:42 p.m. Goldman sent its first genuine order, triggering a surge that electrified the IEX office. As employees watched their screens jitter with unprecedented activity, they leapt from their chairs and began shouting milestones: "Fifteen million!" "Twenty million!" "Fucking Goldman Sachs!" Within 51 minutes before market close, they surpassed the American Stock Exchange in market share.
Brad recognized the significance immediately: "We needed one person to buy in and say, 'You're right.' It means that Goldman Sachs agrees with us." The data revealed the power of trust-Goldman had actually sent more orders the previous day, but on December 19, they entrusted IEX with orders for ten seconds or more. The results were striking: 92% of those orders traded at the midpoint (the fair price) compared to just 17% in Wall Street's dark pools. Their average trade size was twice the market average despite other banks' efforts to undermine them.
IEX demonstrated that the intentionally complicated market could be understood and that a free financial market didn't need kickbacks, payment for order flow, co-location, and unfair advantages. "The backbone of the market is investors coming together to trade," Brad concluded.
The costs of the high-frequency trading infrastructure include market instability, billions collected by intermediaries (essentially a tax on investment), the influence of money on career choices, and perhaps most destructively, the incentive for smart people to further exploit system flaws rather than fix them. Yet Brad's team had shown that even in a system designed to maximize exploitation, a small group of determined individuals could create meaningful change by simply choosing to build something better.
