High-frequency trading
Based on Wikipedia: High-frequency trading
In the time it took you to read this sentence, a high-frequency trading firm could have executed thousands of trades, made millions of dollars, and moved on to the next opportunity. These aren't exaggerations. We're talking about computers that think in microseconds—millionths of a second—competing for advantages so small that light itself becomes too slow.
This is the world of high-frequency trading, or HFT. It's a world where geography matters not because of time zones, but because of how far electricity must travel through fiber optic cables. A world where companies have abandoned fiber optics entirely in favor of microwave towers, shaving precious milliseconds off transmission times. A world so strange that one firm reportedly explored bouncing signals off the moon.
What High-Frequency Trading Actually Is
At its core, high-frequency trading is exactly what it sounds like: using computers to trade financial instruments at extraordinarily high speeds. But speed is just the entry ticket. What makes HFT distinctive is the combination of three elements: sophisticated algorithms, massive transaction volumes, and holding periods measured in fractions of a second.
Traditional investors buy stocks and hold them for months, years, or decades. Day traders might hold positions for hours. High-frequency traders? They're in and out of positions before you could blink. Literally—a human blink takes about 300 milliseconds, and HFT trades can execute in microseconds, which are thousands of times faster.
The profits on any single trade are minuscule. We're talking about fractions of a cent. But when you're executing millions of trades per day, those fractions add up. In 2009, at the peak of HFT profitability, firms were collectively earning an estimated five billion dollars annually. That figure has since declined substantially—by 2012, it had dropped to about $1.25 billion—as the field became more competitive and the easy arbitrage opportunities were squeezed out.
The Speed Arms Race
To understand modern HFT, you need to understand the physics of information.
Light travels at roughly 186,000 miles per second in a vacuum. That's fast enough to circle the Earth more than seven times in a single second. But light through fiber optic cable is slower—about 30% slower, actually, because photons bounce around inside the glass rather than traveling in a straight line.
For high-frequency traders, that 30% slowdown is unacceptable.
Starting around 2011, HFT firms began building networks of microwave towers between major financial centers. Microwaves traveling through air lose less than 1% of their speed compared to light in a vacuum. The difference between microwave and fiber optic transmission between New York and Chicago is just a few milliseconds. But in the HFT world, a few milliseconds is an eternity.
Companies have built microwave networks not just between New York and Chicago, but between London and Frankfurt, routing through Belgium using towers that once belonged to the U.S. military. Some firms have experimented with shortwave radio for even longer distances, though shortwave can carry less information. There have even been discussions about using satellites, and reportedly, investigations into bouncing signals off atmospheric phenomena.
The race has pushed into stranger territory. Firms pay millions of dollars for "co-location"—the right to place their servers physically inside the stock exchange's data center, often within the same room as the exchange's matching engine. Every foot of cable matters. Companies have been known to measure the exact length of their network cables to ensure they aren't even an inch longer than necessary.
How It All Began
High-frequency trading didn't emerge overnight. Its roots trace back to 1983, when NASDAQ introduced purely electronic trading. Before that, trading required human beings—brokers on exchange floors, shouting and gesturing to execute orders. Electronic trading replaced shouting with data packets, opening the door for automation.
At first, computerized trading was merely faster than human trading. By the turn of the millennium, trades executed in several seconds. By 2010, that had compressed to milliseconds and microseconds. The progression wasn't linear; it was exponential.
For years, HFT remained obscure outside the financial industry. Most people had never heard of it. That changed in July 2009, when the New York Times published one of the first mainstream articles exploring this hidden world of algorithmic warfare. Suddenly, the public became aware that their stock markets were dominated not by human judgment, but by competing computer programs.
The growth was remarkable. In the early 2000s, high-frequency trading accounted for less than 10% of equity orders. By 2009, that figure had exploded to 73% of all equity order volume in the United States—despite HFT firms representing just 2% of the roughly 20,000 trading firms in existence.
The Flash Crash: When Speed Becomes Dangerous
On May 6, 2010, something terrifying happened.
In a span of about 36 minutes, the Dow Jones Industrial Average plunged nearly 1,000 points—at the time, its largest intraday point decline in history. Then, almost as quickly as it had fallen, it recovered. Procter & Gamble shares dropped to a penny. Accenture traded for a single cent. Some exchange-traded funds traded at prices that made no mathematical sense.
The event became known as the Flash Crash, and it sent shockwaves through financial regulation.
Subsequent investigations found that both algorithmic and high-frequency traders contributed to the volatility. What happened, essentially, is that the market's "liquidity providers"—the firms that promise to be there to buy when others want to sell—withdrew almost instantaneously when conditions became uncertain. They simply unplugged.
This exposed a fundamental tension at the heart of high-frequency trading. Traditional market makers—firms that continuously offer to buy and sell securities—have regulatory obligations. They're supposed to stay in the market even when things get ugly. HFT firms that perform similar functions have no such requirements. They can and do disappear at the first sign of trouble.
The Flash Crash wasn't a one-time event. Similar mini-crashes have occurred since, reinforcing concerns that our markets have become dependent on systems that can fail in unprecedented ways.
The Strategies: How Computers Make Money in Microseconds
High-frequency trading isn't a single strategy—it's a collection of approaches that share common characteristics. Let's explore the major ones.
Market Making
When you buy a stock, someone has to sell it to you. When you sell, someone has to buy. Market makers are the middlemen who ensure there's always someone on the other side of your trade.
Here's how it works: a market maker continuously offers to buy shares at one price (the "bid") and sell them at a slightly higher price (the "ask" or "offer"). The difference between these prices is called the bid-ask spread, and it represents the market maker's profit margin.
If a stock is bid at $50.00 and offered at $50.02, the market maker is essentially saying: "I'll buy from you for $50.00 or sell to you for $50.02." If they buy from one person and sell to another, they pocket two cents per share.
Traditional market making was dominated by specialist firms with regulatory obligations. High-frequency trading has transformed this business. HFT market makers use sophisticated algorithms to adjust their quotes thousands of times per second, responding to market conditions in real time. They've made spreads narrower—which is good for ordinary investors—but they've also made markets more dependent on computer systems that can vanish in an instant.
Statistical Arbitrage
Arbitrage, in its purest form, is the practice of profiting from price differences in different markets. If gold is selling for $2,000 an ounce in New York and $2,001 in London, an arbitrageur could theoretically buy in New York and sell in London for a guaranteed profit.
Statistical arbitrage is more sophisticated. Instead of looking for absolute price differences, it looks for statistical relationships between securities that have temporarily deviated from their normal patterns.
For example, if two oil companies' stock prices usually move together but suddenly diverge, an algorithm might bet that they'll return to their normal relationship—buying the one that's become relatively cheap and selling the one that's become relatively expensive.
These relationships can involve dozens or hundreds of securities simultaneously. The deviations are tiny and fleeting. They're invisible to human traders but perfectly apparent to computers processing millions of data points per second.
Latency Arbitrage
This is perhaps the most controversial HFT strategy, and it's the one that most resembles what critics call "front-running."
Here's the scenario: a large institutional investor—say, a pension fund—wants to buy a million shares of a stock. That order is too large to execute all at once without moving the market, so it's split into smaller pieces and executed over time.
High-frequency traders can sometimes detect these large orders by analyzing patterns in market data. They notice that someone seems to be systematically buying, and they race ahead to buy shares first, then sell them to the original buyer at a slightly higher price.
Is this legal? It depends on how you obtain the information. Using publicly available market data to infer trading patterns is legal. Getting advance information about orders through special relationships would be illegal front-running.
The line between these is fuzzy, and it's a major source of controversy.
Index Arbitrage
Index funds—like those that track the S&P 500—must buy and sell stocks when the index's composition changes. If a company is added to the S&P 500, every fund that tracks the index must buy its shares. If a company is removed, they must sell.
These changes are announced in advance, creating predictable buying or selling pressure. High-frequency traders can position themselves ahead of this wave, buying before index funds must buy, then selling to them at a profit.
This is legal—the information about index changes is public—but critics argue it represents a tax on index fund investors, whose returns are diminished by traders extracting value from these predictable flows.
News-Based Trading
Computers can read news faster than humans. Much faster.
Automated systems scan headlines from Bloomberg terminals, press releases, social media feeds, and other sources. They identify company names, keywords, and even attempt to understand sentiment. When relevant news breaks, these systems can execute trades before most human traders have finished reading the headline.
This creates a strange dynamic where market prices can move instantaneously in response to news, but humans are still processing what happened. By the time you've understood the implications of an earnings report, the market has already fully incorporated that information.
Quote Stuffing
Not all HFT strategies are benign. Quote stuffing is considered market manipulation and is subject to regulatory penalties.
The technique involves rapidly entering and withdrawing huge numbers of orders—not with the intention of executing them, but to flood exchange systems with data. This can slow down competitors' systems, create artificial confusion in the market, and potentially generate trading opportunities for the quote stuffer.
Distinguishing between legitimate high-frequency activity and manipulative quote stuffing is technically challenging, which is part of why enforcement has been difficult.
The Physics of Trading
Let's pause to appreciate just how extreme the speed competition has become.
Light travels about one foot per nanosecond (a billionth of a second). In the roughly 200 microseconds it takes for a fast HFT trade to execute, light has traveled only about 37 miles. That's barely enough to get from Manhattan to the New Jersey suburbs.
This is why co-location matters so much. The difference between having your server in the exchange's data center versus across the river can be the difference between profit and loss.
Consider the New York to Chicago route, crucial because many financial instruments are traded in both cities. The straight-line distance is about 720 miles. Light in a vacuum covers that in about 3.9 milliseconds. Through fiber optic cable, which takes a longer physical path and slows light down, it's more like 7-8 milliseconds.
HFT firms have spent hundreds of millions of dollars shaving a millisecond or two off that route—building microwave tower networks that take straighter paths and lose less speed to the transmission medium.
When firms invest this much in speed, you know the profits from being first are enormous.
The Case For HFT
Proponents of high-frequency trading make several arguments in its defense.
First, HFT has dramatically narrowed bid-ask spreads. In the 1990s, spreads on major stocks were often 12.5 cents or more. Today, they're frequently a penny or less. This directly benefits ordinary investors, who pay less to buy and receive more when they sell.
Second, HFT has increased market liquidity. There are more buyers when you want to sell and more sellers when you want to buy. Markets function more smoothly as a result.
Third, HFT has made markets more "informationally efficient"—meaning prices more quickly reflect all available information. When news breaks, prices adjust almost instantaneously rather than taking hours or days to reflect new realities.
Fourth, the competition among HFT firms means that the benefits flow to ordinary investors rather than being captured by slow, fat, happy intermediaries. The old specialist system on the New York Stock Exchange was often criticized for allowing those specialists to extract excessive rents from their privileged positions. HFT has disrupted that comfortable arrangement.
The Case Against
Critics have equally compelling arguments.
The Flash Crash revealed that HFT has made markets more fragile. Liquidity can vanish in milliseconds, precisely when it's needed most. The market-making function has been taken over by firms with no obligation to stand by their quotes during periods of stress.
Some HFT strategies, particularly latency arbitrage, appear to extract value from other market participants without providing any corresponding benefit. When high-frequency traders profit by getting ahead of large institutional orders, who pays? Ultimately, it's pension funds, mutual funds, and the ordinary investors whose money those funds manage.
The arms race mentality has led to massive investments in infrastructure—microwave towers, co-location facilities, specialized hardware—that produce no economic value for society at large. These billions of dollars could have been invested in productive enterprises instead of in shaving microseconds off trading times.
The complexity of modern markets, driven largely by HFT, has made them nearly impossible for regulators to monitor effectively. How do you supervise a market where significant events happen faster than human perception?
Finally, the fundamental fairness question: is it right that some market participants have structural advantages—faster connections, co-located servers, microwave networks—that ordinary investors can never hope to match?
The Regulatory Response
Governments have struggled to respond to high-frequency trading.
In September 2013, Italy became the first country to impose a tax specifically targeting HFT. The levy of 0.02% applies to equity transactions lasting less than half a second. The logic is straightforward: if you're holding a position for such a brief period, you're not investing in any meaningful sense, and the activity should be discouraged.
Several European countries have proposed similar measures or outright bans. The debate continues about whether such restrictions would harm market quality or simply level the playing field.
In the United States, regulators have focused more on surveillance and enforcement than on structural changes. The Securities and Exchange Commission (SEC) has brought cases against firms for manipulative practices like quote stuffing, but the fundamental business of high-frequency trading remains legal.
The difficulty is that many HFT activities provide genuine benefits—narrower spreads, deeper liquidity—while potentially harmful activities are hard to distinguish from legitimate ones. A firm rapidly canceling orders might be engaged in illegal quote stuffing, or it might simply be adjusting to changing market conditions. The line is blurry.
The Firms Behind the Speed
Who are these high-frequency traders?
The major American HFT firms include Virtu Financial, Tower Research Capital, IMC, Tradebot, Akuna Capital, and Citadel LLC. These are not household names, and most operate with little public visibility.
Virtu Financial is perhaps the most transparent, having gone public in 2015. In its IPO filings, Virtu disclosed something remarkable: of the 1,238 trading days from 2009 through 2014, the firm had lost money on only one single day. This near-perfect record illustrated both the profitability of HFT and raised questions about whether such consistent profits indicated something other than pure market-making.
Renaissance Technologies, the legendary quantitative hedge fund founded by mathematician Jim Simons, was a pioneer in combining HFT with sophisticated quantitative strategies. The firm's Medallion Fund has produced returns so extraordinary—reportedly averaging over 60% annually before fees for decades—that it accepts only employee money.
The typical HFT firm is structured as a limited liability company owned by a small number of investors. They don't hold significant capital overnight, don't accumulate large positions, and operate very differently from traditional investment firms. Their edge comes from speed and algorithms, not from insight into company fundamentals or long-term economic trends.
Beyond Stocks: HFT in Other Markets
While stocks get most of the attention, high-frequency trading has spread to virtually every electronic market.
In foreign exchange—the world's largest financial market, with daily turnover exceeding $6 trillion—HFT accounts for an estimated 10-15% of volume. In futures markets, the penetration is even deeper: by 2012, HFT accounted for more than 60% of all futures market volume on U.S. exchanges.
Commodity markets, bond markets, and cryptocurrency exchanges have all seen the arrival of high-frequency traders. Wherever there's an electronic market with tight spreads and liquid instruments, HFT follows.
The strategies are similar across markets, though the specific algorithms vary. Statistical relationships between crude oil and refined gasoline futures differ from those between related stocks, but the mathematical principles are analogous.
The Decline of Easy Profits
Ironically, the success of high-frequency trading has undermined its own profitability.
When HFT was novel, there were abundant opportunities for quick profits. Simple arbitrage strategies that a clever programmer could implement generated consistent returns. But as more firms entered the space and algorithms became more sophisticated, the easy money disappeared.
Profits in the U.S. declined from an estimated peak of $5 billion in 2009 to about $1.25 billion by 2012. The decline has likely continued since then. Each new entrant competes away some of the available profit, and the arms race in speed requires ever-larger infrastructure investments.
Many of the algorithms that generate profits today are not fundamentally more clever than their predecessors—they're just faster. The competition has shifted from "who can find the best strategy?" to "who can execute the known strategies quickest?" This is a Red Queen's race: everyone must run faster just to stay in place.
What the Future Holds
High-frequency trading will likely remain a significant force in financial markets for the foreseeable future. But its evolution is uncertain.
The speed arms race is approaching physical limits. You can't transmit information faster than the speed of light, and firms are already operating close to those theoretical boundaries. The next competitive frontiers may involve artificial intelligence and machine learning—algorithms that adapt and learn rather than simply executing predefined strategies faster.
Regulatory pressure may intensify, particularly if another Flash Crash shakes public confidence. The tension between market efficiency and market stability is real, and regulators are still figuring out how to balance these concerns.
The spread of HFT to new asset classes and new geographic markets will continue. As more exchanges become electronic and more countries open their markets to foreign participants, high-frequency traders will follow.
Whether high-frequency trading is ultimately good or bad for financial markets remains hotly debated. The honest answer is probably "both"—it has genuine benefits in terms of tighter spreads and deeper liquidity, but genuine costs in terms of market fragility and fairness concerns. Like many technological revolutions, it has produced winners and losers, and deciding who should win and lose is a political as much as a technical question.
What's certain is that the financial markets of today bear little resemblance to those of even 20 years ago. The shouting crowds on exchange floors have been replaced by humming server racks in nondescript data centers. Human judgment has been supplemented—and often supplanted—by algorithmic decision-making. And the race for speed continues, pushing ever closer to the fundamental limits imposed by the laws of physics.
In the end, high-frequency trading is a mirror reflecting our broader technological moment: a world where computers think faster than humans, where milliseconds matter, and where the old ways of doing things are constantly being disrupted by those who can move quicker. Whether you find that exciting or alarming probably says as much about you as it does about the trading algorithms themselves.