High-frequency trading (HFT) is a form of algorithmic trading that uses sophisticated computer programs to execute large numbers of orders in fractions of a second. It is not a strategy a retail trader runs from a desktop. It is an institutional operation built on speed, scale, and infrastructure that most individual market participants will never access directly, but will encounter in every session they trade.
Understanding what HFT is, how it works, and what it actually does to the markets is practical knowledge for any active trader. It explains why certain executions behave the way they do, where liquidity comes from and where it goes, and why market behavior around large moves can look so chaotic so fast.
What High-Frequency Trading Actually Is
HFT firms are proprietary trading operations. They hold few or no overnight positions. Their entire edge is speed: the ability to analyze market data, identify an opportunity, and send an order in less time than it takes a human to perceive that anything has happened. Execution speeds are measured in milliseconds (thousandths of a second) or microseconds (millionths of a second). A human eye blink takes roughly 400 milliseconds. HFT operates orders of magnitude faster.
The algorithms that power HFT are not looking for fundamental value. They are scanning for temporary price inefficiencies across exchanges, correlated instruments, or order flow patterns that suggest where prices will move in the next fraction of a second. When they find one, they act on it before anyone else can.
The profit on any single trade is tiny, often a fraction of a cent. The volume makes it work. HFT firms execute millions of trades per day. The cumulative take from millions of small edges compounds into substantial revenue at scale. Sharpe ratios at well-run HFT operations have historically run in the high single digits or double digits, far above what traditional buy-and-hold strategies produce, because the exposure per trade is brief and positions are reset constantly.
The Core Strategies
Most HFT activity falls into a small number of strategy categories. The mechanics differ, but the common thread is that all of them depend on being faster than everyone else.
Market Making
The most common HFT approach is electronic market making. The firm continuously quotes both a bid and an ask for a given security, capturing the spread as buyers and sellers transact through it. Traditional market makers were specialist firms with formal obligations to maintain continuous quotes. HFT firms act similarly but carry no such obligation. They post quotes, earn the spread when trades execute against them, and withdraw when conditions turn unfavorable. That distinction matters: a real market maker stays through volatility; an HFT market maker can disappear the moment the risk calculus changes.
Statistical Arbitrage
Statistical arbitrage exploits predictable relationships between correlated securities. Two instruments that historically move together (say, an ETF and the futures contract that tracks the same index) will occasionally drift slightly out of alignment. HFT algorithms detect the divergence and trade both legs simultaneously, locking in the spread before it closes. The profit on any single instance is small. Done thousands of times per day across hundreds of correlated pairs, it adds up.
Latency Arbitrage
Latency arbitrage is more controversial. It targets the small delays that exist between different exchanges and data feeds. If price information reaches one venue 50 microseconds before it reaches another, a firm with the faster connection can act on the stale price at the slower venue before it updates. This is not illegal, but it is genuinely extractive: the profit comes directly from other participants whose information was slower.
News-Based and Event Trading
Automated systems scan electronic news feeds, earnings releases, regulatory filings, and social media for keywords and sentiment signals. When relevant content hits, the algorithm parses it and sends orders before human traders have read the headline. The advantage window is short but decisive. A human trader who sees a headline and then enters an order has already lost several seconds. At HFT speeds, several seconds is an eternity.
Index Arbitrage
When a stock is added to or removed from a major index, funds that track that index are obligated to buy or sell accordingly. HFT firms can anticipate these flows based on publicly available information about index rebalancing schedules, buy ahead of the institutional demand, and sell into it. The pension fund or index fund pays slightly more; the HFT firm pockets the difference.
The Infrastructure Behind It
The strategies above are not executable without infrastructure that most traders will never see. Two elements are central.
Co-location means physically placing servers inside or adjacent to the exchange’s data center. The speed of light is a real constraint. A server 10 miles from an exchange receives market data measurably later than a server housed in the same building. HFT firms pay exchange fees to co-locate their machines as close to the matching engine as possible. The latency difference is measured in microseconds, but at this level of competition, microseconds are the entire margin.
Network transmission between geographic locations is another arms race. Fiber optic cables carry data at roughly 70% of the speed of light. Microwave networks, which HFT firms have built extensively across routes like New York to Chicago and London to Frankfurt, suffer less than 1% speed reduction compared to light in a vacuum. Since at least 2011, firms have invested heavily in microwave tower infrastructure specifically to shave microseconds off intercity data transmission. Some firms have experimented with shortwave radio for even longer distances.
The cost of this infrastructure is enormous. It creates a natural barrier to entry that keeps HFT a domain of well-capitalized institutions, not individual traders.
The Market Share Picture
HFT’s penetration of US equity markets grew rapidly in the early 2000s. By 2009, HFT firms represented roughly 2% of the estimated 20,000 trading firms operating at the time but accounted for approximately 73% of all equity order volume. According to TABB Group estimates, HFT represented about 56% of equity trades in the US by value in 2010. More recent figures indicate HFT continues to account for more than half of US stock trades.
The major firms in the space include Citadel LLC, Virtu Financial, Jane Street Capital, Hudson River Trading, Tower Research Capital, and IMC, among others. These are not household names to most retail traders, but they are counterparties to a significant portion of every trade executed on US exchanges.
Profitability has declined from the industry’s peak. Estimates placed total US HFT profits at roughly $5 billion in 2009. As more firms deployed similar strategies and competition intensified, margins compressed. By 2012, estimates had fallen to approximately $1.25 billion. The strategy space has matured. The pure latency edge is smaller than it was because every serious firm has already optimized the obvious infrastructure.
What HFT Does to Markets
This is where the debate gets substantive. The effects of HFT on market quality are real but contested.
The Liquidity Argument
HFT proponents argue that electronic market makers increase liquidity and narrow bid-ask spreads. This is measurable. A study examining the introduction of HFT fees in Canada found that market-wide bid-ask spreads increased by 13% and retail spreads increased by 9% when fees were imposed on HFT activity. When HFT firms are discouraged from participating, spreads widen. That directly costs retail traders more on every execution.
The Disappearing Liquidity Problem
The counter-argument is that HFT liquidity is not the same as real liquidity. Traditional market makers have obligations to maintain quotes through volatility. HFT firms do not. Their systems are designed to detect adverse conditions and withdraw. The liquidity they provide is real during calm markets and unavailable precisely when it is most needed.
The SEC and CFTC joint report on the 2010 Flash Crash captured this dynamic directly: during the crash, “market makers and other liquidity providers widened their quote spreads, others reduced offered liquidity, and a significant number withdrew completely from the markets.” What appeared to be a deep, liquid market evaporated in minutes because the liquidity was algorithmic and conditional, not committed.
Volatility
Research on HFT’s relationship to volatility is not settled, but the weight of more recent evidence points toward amplification rather than dampening, particularly under stress. Andrew Haldane, then Head of Financial Stability at the Bank of England, found in a 2011 study that intraday volatility had risen most in markets open to HFT, and that HFT algorithms tend to amplify cross-stock correlation when volatility rises. A separate analysis by Dichev, Huang, and Zhou reported that high trading volume “injects an economically substantial layer of volatility above and beyond that based on fundamentals,” estimating that trading-induced volatility accounts for roughly a quarter of total observed stock volatility.
The honest summary is that HFT appears to reduce transaction costs and tighten spreads in normal, calm market conditions. During volatile or declining markets, the evidence suggests it can accelerate dislocations.
The Flash Crash
May 6, 2010, is the most cited example of HFT’s potential to destabilize markets, and it is worth understanding specifically what happened.
A mutual fund, Waddell & Reed Financial, sold $4.1 billion in E-mini S&P 500 futures contracts using an automated execution algorithm that prioritized volume over price. As the selling pressure built, HFT firms were simultaneously selling the same contracts. The SEC and CFTC joint report described the dynamic: high-frequency firms “were also aggressively selling the E-mini contracts,” and then began rapidly buying and reselling contracts to each other in a “hot-potato” volume effect where the same positions changed hands repeatedly without reducing the underlying pressure. The E-mini price dropped 3% in four minutes.
The cascade moved into equities. As automated systems detected the futures dislocation, liquidity in equity markets evaporated. Shares of Procter & Gamble traded as low as a penny. Accenture briefly printed at $0.01. The market recovered most of the losses within 20 minutes, but in that window nearly $1 trillion in market value disappeared temporarily. The incident exposed how fragile the apparent depth of modern markets can be when the computerized participants that supply most of the liquidity decide simultaneously to exit.
The crash was not caused by HFT alone. But HFT firms amplified a bad situation into a near-catastrophic one by withdrawing liquidity and adding selling pressure at exactly the wrong moment.
Regulatory Response and Enforcement
Regulators have responded to HFT with a combination of new rules, fines, and structural changes, though the debate over whether the response has been sufficient continues.
Following the Flash Crash, the SEC imposed mandatory circuit breakers for trading platforms and introduced rules against flash orders. In 2011, new rules required registration of entities conducting substantial algorithmic trading, increasing transparency around who is actually active in the market.
Enforcement has been consistent. In March 2012, regulators fined Getco’s market-making subsidiary $450,000 for failing to maintain proper supervision over its high-frequency trading. In October 2013, Knight Capital was fined $12 million after a software malfunction sent millions of unintended orders into the market over 45 minutes, causing $460 million in losses to the firm before the problem was contained. In September 2014, Tower Research Capital subsidiary Latour Trading agreed to a $16 million SEC penalty, at the time the largest penalty ever for a violation of the net capital rule, after it was found to have systematically underestimated its risk exposure, at times accounting for 9% of all US stock trading while holding insufficient capital.
The CFA Institute’s position is measured: HFT is not inherently manipulative, but the application of the technology can produce manipulative outcomes. The organization supports HFT operating under appropriate regulation, including meaningful oversight of algorithms, controls to catch and correct errors automatically, and prohibitions on unfiltered direct market access that bypasses a broker-dealer’s own risk controls.
What This Means for Active Traders
Individual day traders and swing traders do not compete with HFT firms. The timescales are too different. A momentum trader holding a position for 20 minutes is not playing the same game as an algorithm that exits in 200 microseconds.
But HFT shapes the environment retail traders work in, and understanding that matters for execution.
Spreads in large-cap liquid stocks are tight in part because HFT market makers compete aggressively to capture that spread. That benefits execution on entry and exit. In small-cap or low-float stocks where HFT participation is lower, spreads widen and execution quality deteriorates accordingly. A stock with a $0.05 spread costs a trader $500 per 10,000 shares round-trip just in spread friction, before commissions.
The behavior of Level 2 during fast markets reflects HFT activity. When a catalyst hits and a stock makes a sharp move, the visible bids and offers on Level 2 can disappear and reform at completely different price levels within seconds. What appeared to be a 2,000-share offer at $14.20 may have been a quote from an HFT market maker that canceled the instant the stock started moving. That is not manipulation; it is an algorithm doing exactly what it is designed to do. Understanding this prevents over-reliance on Level 2 as a static picture of supply and demand.
Slippage on market orders in fast-moving stocks is partly a function of HFT. In a normal market, HFT market makers provide tight quotes and execute against market orders efficiently. When momentum accelerates and HFT systems detect adverse selection, meaning they realize they are consistently on the wrong side of informed order flow, they widen quotes or pull them entirely. That is when slippage spikes. Limit orders are the practical response.
The Bottom Line
High-frequency trading is a permanent feature of modern equity markets. It is responsible for a large share of daily volume, narrows spreads in liquid stocks under normal conditions, and makes markets function more efficiently in the mechanical sense of connecting buyers and sellers quickly. It also withdraws at the worst possible moments, can amplify volatility, and operates with advantages in speed and information that individual traders cannot replicate.
The practical takeaway is not that HFT is good or bad, but that it changes what market data means and how executions behave. Level 2 depth is conditional. Apparent liquidity can disappear. Fast markets create slippage risks that calm markets do not. Knowing the mechanism behind these behaviors does not eliminate them, but it removes the confusion of treating market data as something more reliable than it is.