Speed, not skill, became the way to skim stock trades

1

Mountains for milliseconds

An industry spent roughly $300 million blasting a straighter fiber path because a few milliseconds of lead could be sold, repeatedly, as pure edge.

Lewis opens on the claim that speed itself, not insight, had become the dominant advantage. The market was no longer one place; it was a scatter of exchanges linked by cables, and the shortest cable won.

Construction crews cut a near-straight line through the Alleghenies between Chicago and New Jersey exchanges. Ridges were tunneled rather than routed around, because every bend cost time that traders would pay to avoid.

The line was sold to banks and trading firms for large annual fees. Each buyer’s choice was rational; in aggregate the race produced no new information, only a faster mirror of prices that already existed a few milliseconds away.

Ordinary investors never heard of the project, yet its existence quietly rewrote how professional trading worked. Control of the physical path became a reproducible source of profit.

In practice: Spread Networks spent roughly $300 million to cut the round-trip data time from about 16 milliseconds to 13, then leased the advantage to anyone who needed to arrive first.

2

The order that arrived late

When pieces of a large order hit separated exchanges at slightly different times, faster firms saw the first fill and raced ahead to buy elsewhere and resell higher.

Brad Katsuyama at RBC watched prices move against him in the gap between click and execution whenever he tried to trade a sizable position across multiple venues at once. It felt as if the market already knew his intent.

He first assumed an internal bug and spent months eliminating desk and system explanations. The real cause was geography: an order sent “simultaneously” still reaches nearer exchanges before farther ones.

High-frequency firms colocated at or near every venue, with faster links than ordinary brokers, could detect the first arrival, infer the rest of the order was inbound, and step in front on the remaining exchanges. The edge was speed, not secret research.

That mechanism turned the earlier cable race into a concrete extraction from institutional desks—and, by extension, from the pension and mutual-fund money those desks represented.

In practice: Katsuyama’s team saw the pattern repeat on large stock positions until they proved the lag, not their own software, was moving the price.

3

One market becomes many

Regulation NMS, finalized in 2005 to guarantee the best price everywhere, multiplied venues and made every order’s path mechanically predictable to anyone watching the first hop.

Lewis traces the structure to SEC rules meant to protect investors by forcing brokers to honor the best displayed quote across the market. The unintended result was a jump from a handful of exchanges to more than a dozen public ones plus dozens of private dark pools, each of which had to be checked.

Because those venues sit at real distances, true simultaneous checking is impossible; a small lag always remains. Best-price routing also made the sequence of an order legible once it began to move.

Exchanges earned fees from colocation and from selling faster private data feeds ahead of the public ones. They also rolled out specialized order types whose practical effect, critics said, was to let speed firms sniff incoming interest before others could react.

Fragmentation was sold as competition and protection. Lewis argues it created the exact seams that turned distance and delay into an extractive strategy the rule-writers never anticipated.

In practice: After the 2005 rules took hold, firms poured capital into servers inside exchange data centers and into private networks built solely to exploit the new lags.

4

Brokers paid to sell flow

Dark pools and payment for order flow turned intermediaries into sellers of their customers’ information, quietly aligning pay against the investors they served.

Dark pools were marketed as private matching venues where large orders could hide from public view and avoid moving the market. Lewis shows many of them sold high-frequency firms privileged access to see the flow inside, recreating the information edge the pools were supposed to erase.

Separately, retail brokers sold ordinary customers’ orders to wholesale trading firms instead of sending them straight to an exchange. The broker collected a fee; the wholesaler traded against the flow and kept the difference after giving the customer a price only marginally better than the public quote.

Banks pitched institutional clients on dark-pool safety from predators while, in the same period, selling access to those same predators. No single actor had to be cartoonishly corrupt.

Each layer—exchange, dark pool, retail broker—followed incentives the fee structures and rules had created. The system became extractive through ordinary rational choices.

In practice: Lewis describes banks selling dark-pool safety to institutions while high-frequency firms paid for the right to see order flow inside those same pools.

5

Thor closes the window

Staggering the send times so every piece of an order landed on every exchange together erased the gap predators needed, and clients saw better prices as proof.

Once the lag was understood, Katsuyama’s group, including Rob Park, built a tool that deliberately delayed some legs of an order. The goal was simple: identical arrival times everywhere, no first hop for anyone to observe.

When RBC offered the tool to institutional clients, execution improved in measurable ways. The losses that had looked like bad luck or market noise showed up as systematic and removable.

Institutional politics followed. Banks that ran their own speed desks or sold speed and data to others had little reason to celebrate a fix that shrank that profit pool. Some clients hesitated too, because enthusiastic adoption meant admitting the market had been broken and that parts of their own firms had benefited.

A single broker tool could not rewrite surrounding incentives. Money still preferred the status quo, which pushed the team toward a more radical step: an exchange that did not rely on anyone choosing to be fair.

In practice: Institutional clients using Thor received meaningfully better prices than they had on the same kinds of orders without it, evidence the earlier slippage was real.

6

Fairness in a coil of fiber

IEX’s 350-microsecond “magic shoebox” delay, applied equally to every order, was long enough to outlast the fastest race-ahead reaction and short enough that humans never noticed.

Katsuyama left RBC to found IEX around a physical rather than moral solution. A coil of extra fiber added a fixed delay to every inbound order, calibrated to exceed the reaction time of the quickest high-frequency links.

Because the delay was symmetric, colocation and private cables lost their edge inside the venue. IEX also refused the specialized order types and rebate designs Lewis ties to selling speed advantages, favoring a simpler fee model.

Staff had to be recruited from firms that still profited from the old architecture, and large asset managers had to be persuaded to send real flow to an unproven exchange. Early resistance from incumbent exchanges and banks underscored how much the prior design had been worth to them.

The working venue became a proof of concept: markets did not require exploitable speed differentials to function, attract volume, or improve execution.

In practice: The magic shoebox imposed exactly 350 microseconds of delay on every participant, neutralizing the lead that private cables and colocation had previously bought.

7

Volume follows the big funds

Even a fair exchange stays irrelevant until mutual funds and pension funds that control most flow stop taking conflicted routing advice from banks that profit from the old paths.

Technology alone did not finish the job. Without liquidity, institutional traders cannot move size, so IEX had to win the order flow of the managers who hold ordinary savers’ retirement money.

The team pitched heads of trading at major mutual-fund and pension organizations directly, arguing that the hidden costs Katsuyama first measured would fall if orders came to a venue without speed races. Progress was slow.

Many of those institutions still leaned on their banks for execution guidance, and those banks had clear reasons to keep flow inside arrangements they already monetized. Public criticism of the same banks raised reputational risk for a startup that still needed industry cooperation.

Reform therefore meant fighting inertia and misaligned advice, not only writing better code. Structural problems that arise from incentives do not dissolve when one new venue appears.

In practice: Katsuyama’s group spent long stretches winning commitments one large asset manager at a time while the banks those managers trusted still preferred the prior routing map.

8

The fight goes public

Senate testimony and the 2014 publication wave reframed microsecond front-running as a question of national fairness, forcing a defense Lewis finds incomplete on who actually pays.

The narrative peaks when the findings leave trading floors. Katsuyama testifies before a U.S. Senate subcommittee and debates an exchange representative on television; the book’s release coincides with announced FBI and New York Attorney General attention to high-frequency practices.

Defenders answer that high-frequency trading adds liquidity and narrows bid-ask spreads, lowering explicit costs for everyone including retail. Lewis puts that case on the page, then presses the distributional point: average spreads can tighten while fast, informed participants still extract value from slower large orders of the kind Katsuyama first watched slip.

Regulators, lawmakers, and financial media suddenly had to discuss mechanics most had never aired in public. The obscurity itself was gone.

At the time of writing, deep rewrites of Reg NMS and exchange fee structures remained largely unfinished. Visibility was the partial victory.

In practice: Around the 2014 release, Katsuyama’s congressional appearance and a wave of coverage pushed inquiries and industry rebuttals into the open.