Introduction
You likely picture the stock market as a bustling room of shouting traders. That traditional trading floor vanished long ago, replaced by hidden computer servers inside windowless suburban data centers.
High-frequency trading firms now dominate finance using invisible algorithms that operate in milliseconds. When Sergey Aleynikov copied computer code from Goldman Sachs, he triggered an FBI arrest that revealed a much larger financial mystery. You discover that modern markets are no longer fair and open marketplaces for ordinary investors.
Michael Lewis exposes how electronic exchanges allowed predatory traders to exploit institutional orders. The book follows a determined group of outsiders who traced this hidden market tax across fiber-optic cables. You learn how these pioneers built a new, transparent exchange to reclaim market fairness.
The book answers several key questions about modern finance:
- How do high-frequency traders see your orders before they arrive?
- Why did Wall Street banks ignore this predatory behavior?
- What physical realities dictate the speed of electronic trading?
- How can you design an exchange that stops front-running?
1. The Mechanics of High-Frequency Trading
High-frequency trading firms make immense profits by exploiting minuscule price discrepancies between geographically separated stock and futures markets.
Over the decade leading up to the book, human traders in trading pits were entirely replaced by computers operating inside heavily guarded black-box data centers in New Jersey and Chicago. The public perception of markets is outdated, relying on cable TV tickers and pit traders. Modern stock trading occurs in black boxes where ticker feeds capture only a tiny fraction of activity and even experts cannot fully explain the precise timing of events.
Michael Lewis notes how the 1987 Salomon Brothers crash exposed insider ignorance, prompting regulatory rules that encouraged computers to replace humans. On October 19, 1987, the entire U.S. stock market dropped by 22.61 percent.
The arrest of a former Goldman Sachs programmer for taking computer code highlighted the extreme secrecy and massive financial stakes surrounding high-frequency trading. Sergey Aleynikov was arrested on a flight from Chicago to Newark on July 3, 2009. Prosecutors argued his code could be used to manipulate markets in unfair ways. This mystery sparked the author's investigation into high-frequency trading.
Building the Line
Traders make immense profits by exploiting minuscule price discrepancies between futures contracts in Chicago and individual stock prices in New Jersey. Dan Spivey realized that thousands of times a day, prices between the Chicago Mercantile Exchange and New Jersey data centers fell out of whack. Capturing these arbitrage profits required extreme speed to execute trades in both locations simultaneously.
Traders searched for a Verizon route known as the Gold Route that accidentally achieved a faster transit time. Theoretical calculations show it should take 12 milliseconds to send an order from Chicago to New York and back based on the speed of light in fiber. Telecom carriers previously offered speeds of 17 milliseconds, 16 milliseconds, and 14.65 milliseconds before the new line.
To achieve maximum trading speed, a private company undertook a massive engineering project to dig an uncompromisingly straight underground fiber-optic cable straight through the Allegheny Mountains. Existing telecom routes followed railroads and highways, zigzagging around the Allegheny Mountains and causing delays. Dan Spivey and construction expert Steve Williams planned the straightest path allowed by law, cutting over a hundred miles off existing routes.
Spivey hired Jim Barksdale to fund the tunnel project, disguised construction under shell companies like Northeastern ITS and Job 8, and formed Spread Networks.
The construction of the direct fiber line faced severe geographic, technical, and social roadblocks that threatened to delay its completion. Construction crews struggled to bore under rivers like the Calumet and Susquehanna, cut through hard limestone blue rock in Pennsylvania, and negotiate easements with suspicious local property owners.
The project stalled in Sunbury when the owner of Weis Markets refused to grant an easement because the cable passed too close to his ice cream manufacturing plant. Overcoming these obstacles required renting the only drill in the world capable of boring under the Susquehanna River at a cost of $2 million, while routing around Sunbury added an extra delay of four microseconds.
The final length of the completed Spread Networks line reached eight hundred twenty-seven miles, sheltered inside a one-and-a-half-inch-wide hard black plastic tube containing four hundred hair-thin strands of glass.
Selling the Speed
Spread Networks priced access to its ultra-fast fiber line at premium rates, exploiting the massive value of speed to high-frequency trading firms and Wall Street banks. Consultant Larry Tabb estimated that a single Wall Street bank exploiting price discrepancies could make $20 billion annually, with 400 firms vying for the profits. Spread Networks decided to limit access and charge millions for upfront leases.
Hudson River Trading bought into the line through a Dutch auction, while big banks like Morgan Stanley and Goldman Sachs negotiated contracts with strict anti-sharing restrictions and non-disclosure agreements.
Spread Networks charged a monthly lease price of $300,000 a month, with a five-year lease costing $10.6 million paid in advance. The total upfront cost per trader reached $14 million including amplifiers and fees, yielding a grand total of $2.8 billion in upfront revenue if all 200 slots were sold.
2. Brad Katsuyama and the Hidden Market Tax
RBC purchased Carlin Financial for $100 million in late 2006 to gain a foothold in electronic trading. The integration brought abrasive personalities like founder Jeremy Frommer and a group of day traders whose behavior clashed completely with RBC's polite culture.
Jeremy Frommer walked around the office swinging a baseball bat, rented out the Marquee nightclub for a lavish Christmas party, and bragged to business students about flying first class while his friends flew in coach.
By 2007, when Brad tried to buy large blocks of stock based on quotes shown on his screens, the offers would instantly vanish and prices would move against him. This made it impossible to price large investor orders without suffering massive losses.
When a big investor asked Brad to sell 5 million shares of Solectron, Brad's attempts to access public market liquidity triggered a minicollapse in the stock price, causing him to lose a small fortune.
Prompted by the SEC, exchanges transitioned from member-owned utilities to for-profit corporations in 2005. They proliferated into 13 different public exchanges mostly located in northern New Jersey and replaced human specialists with server stacks and complex maker-taker pricing models.
Market fragmentation and disparate exchange speeds created a hidden tax that allowed predatory traders to front-run institutional orders.
When Brad tried to direct RBC's orders to the BATS exchange to collect payment as a taker, the market vanished and trading losses hemorrhaged even further.
Thor the Hammer
Through rigorous science experiments using RBC trading capital, Brad's team discovered that orders sent to multiple scattered exchanges arrived at slightly different microsecond intervals. Faster actors could intercept and front-run the slower orders.
When Rob Park built a program to artificially delay orders heading to faster exchanges so they arrived simultaneously with slower ones, the trading screens suddenly turned green and orders were successfully filled.
Programmed by Allen Zhang and championed by Rob Park and Brad, Thor built intentional delays into stock exchange orders so they struck all disparate exchange matching engines at the exact same moment. This allowed traders to execute large blocks without being front-run.
After testing Thor on internal RBC trades, traders began shouting "Thor it!" as the tool successfully enabled them to buy stock and restore trust in their screens.
By calculating the fractional savings achieved when bypassing high-frequency front-running via Thor, Brad's team estimated that the modern market structure extracted a small fraction of a percent on every transaction. This scaled up to staggering sums across the entire U.S. stock market.
Brad demonstrated to Mike Gitlin of T. Rowe Price that a fraction of a percent tax applied to the $225 billion daily U.S. stock market volume amounted to over $160 million extracted daily.
3. Ronan Ryan and the Physical Reality of Trading
You might assume that digital finance erased geography entirely. In reality, physical distance and fiber routes heavily dictated speed and market advantage.
Physical distance, fiber-optic routes, and broker routing strategies heavily dictate speed and market advantage.
Ronan Ryan understood how data traveled across physical infrastructure like copper circuits, fiber-optic cables, and routers. He eventually became an expert in co-locating computer servers close to exchange matching engines to minimize latency.
Ronan moved a Kansas City hedge fund's computers to Radianz's data center in Nutley, New Jersey, reducing their latency time from 43 milliseconds to 3.8 milliseconds. High-frequency traders spent enormous sums to shave microseconds off travel times by positioning their servers as close as possible to exchange pipes and securing straighter fiber paths. Traders at Radianz data centers paid to move their machines to the exact spot closest to the exit pipe and requested straight fiber cables instead of looping ones. From late 2005 to the end of 2008, Radianz billed high-frequency traders nearly $80 million for setting up near exchange matching engines.
High-frequency trading firms went to extreme lengths to conceal their hardware improvements and strategies from competitors. One HFT firm wrapped its servers in wire gauze to hide blinking lights, while another left old Toys "R" Us store logos on a computer cage to disguise its advantageous proximity to the matching engine. Brad Katsuyama hired Ronan Ryan at RBC to bridge the gap between financial trading and telecom technology. Brad hired Ronan after a brief interview, paying him $125,000, which was a third of what Ronan made selling speed to high-frequency traders. Ronan had earned $486,000 in 2007.
Telecom Expertise
By obtaining proprietary telecom maps of New Jersey, Ronan helped RBC understand why their trading tool Thor was inconsistent. Ronan's maps revealed the exact, convoluted routes fiber cables took from lower Manhattan to various New Jersey exchanges, explaining why orders arrived at different times and allowed HFTs to front-run them. Ronan unrolled oversized maps of New Jersey showing how telecom networks carved up the state, shocking RBC's network support team.
Wall Street brokers used sequential cost-effective routers that inadvertently exposed client orders to high-frequency predators. A sequential cost-effective router would send 100 shares to BATS to collect a rebate, causing the remaining 100,000 shares on other exchanges to vanish into the paws of high-frequency traders.
The Investor Awakening
By openly explaining market mechanics to institutional investors, Brad and Ronan revolutionized RBC's standing and sparked a broader investor awakening. Brad and Ronan bypassed formal PowerPoint presentations, sitting down with hundreds of major money managers to explain how their trades were being exploited, which drastically shifted institutional awareness.
A hedge fund president watched the price of an ETF pop on his Bloomberg terminal the exact moment he clicked buy on his online brokerage account, executed by Citadel Derivatives. By the end of 2010, Brad and Ronan had met with 500 professional stock market investors. RBC's low rank of 19 in Greenwich Associates stock market rankings in 2009 jumped to 1 at the end of 2010 after six months of Thor.
The May 2010 flash crash exposed the inadequacy of regulatory data and heightened institutional suspicions about market fairness. Regulators blamed the flash crash on a single mistakenly placed sell order, but their reports relied on seconds and minutes rather than the microseconds at which modern markets actually operated. During the flash crash at 2:45 on May 6, 2010, the market fell 600 points in minutes, with Procter & Gamble shares trading as low as a penny and as high as $100,000. The official SEC report contained 87 hits for the word minute, 63 hits for the word second, 4 hits for the word millisecond, and 0 hits for the word microsecond.
The involvement of major Wall Street banks in high-speed tunneling projects revealed that market exploitation was an industry-wide, systemic issue. Spread Networks built an ultra-fast fiber line between Chicago and New Jersey for millions of dollars upfront, enabling HFT firms and major banks to front-run trades across financial centers. Spread Networks charged an upfront cost of $10.6 million for five years of service, and Ronan was offered a $12,000 commission per line sold.
4. Tracking the Predator
John Schwall investigated electronic trading systems after experiencing corporate betrayal during Bank of America's acquisition of Merrill Lynch. He mapped out dark pool organizations using LinkedIn and public records.
Wall Street banks and regulators are structurally complicit in high-frequency trading predation, turning the market into a risk-free extraction machine.
John Schwall built FBI-style organizational charts of Credit Suisse's dark pool division, matching résumés and former employees like those from Bernie Madoff's firm to show that banks employed vast high-frequency trading talent. More than a hundred employees at too-big-to-fail banks were interviewed by Brad Katsuyama, and thirty-five were eventually hired by him. John Schwall's desk was located on the eighty-first floor of the North Tower of the World Trade Center, and he had joined Banc of America Securities in the 1990s as a top 1 percent employee in reviews.
Reg NMS and the SIP
Passed in 2005 and implemented in 2007, Regulation National Market System required brokers to route orders to the exchange offering the National Best Bid and Offer calculated by the Securities Information Processor. Because the public SIP feed was technologically slow, high-frequency traders built faster private feeds to see market price changes milliseconds ahead of ordinary investors. This created a twenty-five milliseconds gap between high-frequency traders' view and ordinary investors' view.
Researchers at the University of California, Berkeley showed in 2013 that the SIP price of Apple stock and the price seen by faster traders differed 55,000 times in a single day. High-frequency traders' share of total U.S. stock market trading surged from 26 percent in 2006 to 52 percent in 2008. The total number of trades rose from 10 million per year in 2006 to 20 million per year in 2009. These shifts followed a $241 million fine paid by specialists on the floor of the old New York Stock Exchange in 2004.
The Illusion of Liquidity
Rich Gates created TFS Capital in 1997 to manage $2 billion for 35,000 small investors while earning 10 percent per year returns. His Phoenix Market Neutral Fund lost 0.09 percent a year over a decade. He tested for predatory behavior by placing a mid-market buy order in a dark pool for a lightly traded stock and immediately sending a sell order to a public exchange.
Rich Gates tested dark pools in 2010 and found that Goldman Sachs's Sigma X pool ripped him off more than half the time, and later tests confirmed similar predatory behavior in Credit Suisse's Crossfinder and on the BATS exchange.
High-frequency firms go home flat every night without taking market risk. By 2008, high-frequency trading firms made 65 percent of public stock market trades. By 2011, 64 percent of all stock market trades occurred off public exchanges in dark pools. Price volatility was nearly 40 percent higher between 2010 and 2013 compared to 2004 and 2006. Reaching a seat at Citadel required five badge swipes.
When Brad Katsuyama presented Thor to SEC staffers, some officials defended high-frequency traders, arguing it was unfair to force them to honor quoted prices. More than two hundred SEC staffers since 2007 left government jobs to work for high-frequency trading firms or lobbying firms. In June 2010, the associate director of the SEC's Division of Trading and Markets, Elizabeth King, quit the SEC to work for high-frequency trading firm Getco.
A contact at Citadel told Ronan that Thor was genius and unstoppable, but noted it only accounted for two percent of the market, prompting Rob Park and Brad Katsuyama to conceptualize a new stock exchange. Nine big Wall Street banks controlled nearly 70 percent of all stock market orders. These top nine banks were ranked by 2011 market share as Credit Suisse, Morgan Stanley, Bank of America, Merrill Lynch, Goldman Sachs, J.P. Morgan, Barclays, UBS, Citi, and Deutsche Bank.
5. Building IEX
Brad Katsuyama and key members of his team resigned from the Royal Bank of Canada to launch a new, transparent stock exchange. Ronan returned from visiting his parents in Ireland, where he debated quitting his $910,000-a-year job, before returning to New York to find Brad's resignation message waiting on his BlackBerry.
To eliminate systemic predation, Brad Katsuyama and his team resigned from RBC to launch IEX, a transparent and independent stock exchange.
Large institutional investors urged Brad Katsuyama to fix the stock market but hesitated to provide the startup capital required to build IEX. During pitch meetings, eight out of ten investors asked Brad why he was attacking a system that made him rich, questioning if he was playing Robin Hood.
To design an exchange that could not be gamed, the founders hired national puzzle champions who treated the market as a cryptography puzzle. Dan Aisen won the Microsoft College Puzzle Challenge, leading Brad Katsuyama to nickname him the Puzzle Master.
The Speed Solution
IEX used a coiled length of fiber optic cable to create a physical distance delay that neutralized high-frequency trading speed advantages. To eliminate the advantages of speed without placing the exchange in a distant remote location, IEX coiled 38 miles of fiber optic cable inside a shoebox-sized compartment between data centers.
- Charge both sides of any trade an equal amount of nine one-hundredths of a cent per share.
- Allow just three order types: market, limit, and Mid-Point Peg.
Major Wall Street banks actively undermined IEX by delaying connections, spreading rumors, and failing to route customer orders as instructed. UBS and other big banks failed to route orders to IEX despite explicit instructions from mutual fund managers, following client requests at most ten percent of the time.
Goldman Sachs Turning Point
Goldman Sachs partners Brian Levine and Ron Morgan shifted the firm's approach by taking IEX seriously and eventually routing honest order flow to the new exchange. On December 19, 2013, Goldman Sachs began sending unconstrained order flow to IEX, causing trading volume to surge past 40 million shares in fifty-one minutes and validating the exchange's model.
10 Key Ideas
- Human trading floors vanished as computers inside New Jersey and Chicago data centers replaced traditional methods, sparking intense secrecy after a programmer was arrested for taking Goldman Sachs code.
- Traders spent millions of dollars to build an uncompromisingly straight underground fiber-optic cable through the Allegheny Mountains to exploit minuscule price discrepancies between Chicago and New Jersey markets.
- Fragmented public exchanges and disparate exchange speeds created a hidden tax that allowed predatory high-frequency traders to front-run institutional investor orders.
- Physical distance, fiber-optic routes, and broker routing strategies heavily dictated speed and market advantage, leading firms to co-locate servers and hide their hardware.
- Wall Street brokers used sequential cost-effective routers that inadvertently exposed client orders to high-frequency predators who profited risk-free.
- Wall Street banks and regulators were structurally complicit in high-frequency trading predation, turning the market into a risk-free extraction machine while regulators captured by industry left for private firms.
- High-frequency firms went home flat every night without taking market risk while dark pools handled a massive share of trades with predatory behavior confirmed by testing.
- Brad Katsuyama's team built Thor to introduce intentional delays into stock exchange orders so they struck all disparate exchange matching engines at the exact same moment.
- To eliminate systemic predation, Brad Katsuyama and his team resigned from RBC to launch IEX, a transparent and independent stock exchange.
- IEX used a coiled length of fiber-optic cable to create a physical distance delay that neutralized high-frequency trading speed advantages until Goldman Sachs finally began routing honest order flow to the new exchange.