High-Frequency Trading: What It Is, Its Strategies, Benefits and Risks

High-frequency trading (HFT) is automated trading by firms that use very fast computers and connections to send, change and cancel large numbers of orders. There is no single legal definition. In its 2010 concept release on equity market structure, the SEC listed five characteristics often attributed to HFT firms: "extraordinarily high-speed and sophisticated computer programs for generating, routing, and executing orders"; co-location and individual data feeds "to minimize network and other types of latencies"; "very short time-frames for establishing and liquidating positions"; "numerous orders that are cancelled shortly after submission"; and "ending the trading day in as close to a flat position as possible" (SEC Release No. 34-61358, Federal Register, January 21, 2010).

How big is it?

Nobody publishes an official share. In the same 2010 release, the SEC wrote that estimates of HFT volume in U.S. equities "vary widely, though they typically are 50% of total volume or higher," and that "by any measure, HFT is a dominant component of the current market structure." It is a 2010 estimate, not a current measurement; we found no later official figure.

HFT strategies

The SEC release sorts the strategies of the proprietary firms that trade this way into four groups. They differ in whether they supply liquidity or take it.

• Passive market making. Post buy and sell limit orders on both sides of the book, earn the spread between them and the rebates some exchanges pay for resting orders. The release notes that layering the book this way can bring cancellation rates of 90% or more, with orders lasting "often of a second or less." See bid and ask for how the spread works.
• Arbitrage. Trade the gap between prices that should move together, for example an ETF and the stocks in its basket, or three currency pairs in triangular arbitrage, as soon as it opens.
• Structural. Exploit weak points of the market or of other participants, for example by using the fastest data to find traders still offering executions at stale prices.
• Directional. Take a short-term view on price, including strategies the release calls order anticipation (detecting a large buyer or seller and trading ahead) and momentum ignition (a burst of orders and trades meant to set off a rapid move, after taking a position early to sell into it). The SEC raised these as concerns to examine, not as findings about specific firms.

Speed: the arms race

Speed is measured in millionths of a second. The 2010 release cites exchange latencies of 320 microseconds for BATS in June 2009 and 294 microseconds for Nasdaq in December 2009. The race has kept going since.

Budish, Cramton and Shim (2015) used millisecond data from exchanges and found that correlations between closely related prices "completely break down" at those horizons, which creates "obvious mechanical arbitrage opportunities," and that competition "has not affected the size or frequency of the arbitrage opportunities, it has only raised the bar for how fast one has to be to capture them." They call the arms race "a symptom of flawed market design" and propose batch auctions run, for example, every tenth of a second (Budish, Cramton and Shim, Quarterly Journal of Economics, 2015).

The race in its simplest form: a price moves on one market, and the quotes that depended on it are now stale. The market maker who posted them sends a cancel; faster traders send orders to take them. The exchange serves messages in the order they arrive, so the slower message loses.

Impact on liquidity

An early causal study is on algorithmic trading in general, of which HFT is the fastest part. When the New York Stock Exchange automated its quote dissemination in 2003, algorithmic trading rose. Hendershott, Jones and Menkveld (2011) used that change to measure its effect and found that, "for large stocks in particular," it narrowed spreads, reduced adverse selection and made quotes more informative (Hendershott, Jones and Menkveld, Journal of Finance, 2011).

The two lines of research are not in conflict. Narrow spreads help anyone who trades. Latency races are a cost of the same markets, paid mostly by the slower side of each race. Both can be true at once.

Impact on volatility: the Flash Crash of May 6, 2010

The CFTC and SEC staff report on May 6, 2010, published on September 30, 2010, describes "unusually high volatility and thinning liquidity" that afternoon before the trigger. At 2:32 p.m., a mutual fund complex began an automated program to sell 75,000 E-mini S&P 500 futures contracts, about $4.1 billion, as a hedge. The order was intended, so the day was not a fat finger trade. As prices fell, high-frequency traders bought and resold contracts to each other: between 2:45:13 and 2:45:27 p.m. they traded over 27,000 contracts, about 49% of volume, while adding only about 200 contracts to their net position. Buy-side depth in the E-mini fell to about $58 million, "less than 1% of its depth from that morning's level" (CFTC and SEC staff report, September 30, 2010).

The staff report also gives the depth of the E-mini order book that day, on the buy side, in dollars: nearly $6 billion in the early morning, $2.65 billion by 2:30 p.m., and about $58 million between 2:45:13 and 2:45:27 p.m. At 2:45:28 p.m. the exchange paused E-mini trading for five seconds; when it resumed at 2:45:33 p.m., prices stabilized and began to recover.

In stocks, many liquidity providers widened their quotes or withdrew. Between 2:40 and 3:00 p.m., over 20,000 trades in more than 300 securities were executed at prices 60% or more away from their 2:40 p.m. prices, some at a penny or less; the exchanges and FINRA canceled them after the close. Kirilenko, Kyle, Samadi and Tuzun (2017), using audit-trail data for the E-mini, found that the trading pattern of the most active intraday intermediaries, which they classify as high-frequency traders, "did not change when prices fell during the Flash Crash" (Kirilenko, Kyle, Samadi and Tuzun, Journal of Finance, 2017).

Benefits and risks

What it means for a retail trader

You cannot compete on speed, and a strategy that holds positions for minutes or days does not need to. HFT matters to you through execution: the spread you pay, how fast quotes change on screen, and what happens to a market order when the book thins out.

• Use limit orders when a fill price matters. They cap the price you pay or accept.
• Read the order book with caution. Orders that are canceled within a second can disappear before you reach them. See level 2 market data.
• Hold for longer than the race. Races last millionths of a second; a trade held for days is not playing that game.
• Know where your order goes. Some trading happens off exchange, as in dark pools.

The scale gap is wide. In 2010 the SEC cited average exchange latencies, the time to accept, process and acknowledge or fill an order, of about 300 millionths of a second; in the study published in 2022, the modal race on FTSE 100 stocks lasted 5 to 10 millionths. A trade held for one minute lasts 60 million millionths of a second.

High-frequency trading in forex

Most of the research above is on stocks and stock index futures. For currencies, the Markets Committee of the Bank for International Settlements wrote in September 2011 that high-frequency trading "has increased its presence in the foreign exchange (FX) market in recent years" and that "a discussion is emerging about its benefits and risks" (BIS, High-frequency trading in the foreign exchange market, 2011). We found no official figure for its share of FX volume. The strategies are the same in kind: market making on electronic platforms and arbitrage between related prices, such as the three-pair loops of triangular arbitrage.

A "high-frequency forex strategy" sold to retail traders runs on a retail platform. Without the co-location and direct data feeds the SEC lists among the traits of HFT firms, it is not competing in races that last millionths of a second; judge it like any short-term strategy, on its costs per trade and on results tested over a long period.

More on how orders meet in the market is in trading concepts.