In very broad terms, high-frequency trading (also known as HFT) refers to the buying and selling of stocks at extremely fast speeds with the help of powerful computers. Using complex algorithms, these computers can scan dozens of public and private marketplaces simultaneously, execute millions of orders a second, and alter strategies in a matter of milliseconds. In the U.S., high-frequency trading firms represent 2.0% of the approximately 20,000 firms operating today, but account for 73.0% of all equity trading volume.
As an example, let's assume that a buyer wants to buy 100,000 shares of INTL. The market price of an INTL share is $26.10, but the buyer's limit price is $26.40. In other words, the buyer is willing to pay up to $26.40 for each share of INTL or $0.30 more than its current price.
Via flash orders from NASDAQ, high-frequency trading firms get a peek at these orders for 30 milliseconds before they are shown to everyone else. Having detecting a demand for INTL shares, the computers at these firms then start issuing small immediate or cancel (IOC) orders at specific levels above the current price of INTL shares. If the first sell order at $26.15 is accepted by the buyer, another sell order at $26.20 is issued, and so on.
This continues until a sell order at $26.45 is issued. Because the buyer's limit price is $26.40, the sell order at $26.45 is rejected. At this stage, the firms' computers flood the buyer with sell orders at $26.39, causing most of the company's order of 100,000 INTL shares to be filled at $0.29 cents above market price.
Under normal circumstances, a buyer would see the sell order at $26.15 and might subsequently drop the limit price on his/her order. However, high-frequency trading computers are so fast that unless the buyer owned comparable machines, he/she would have no chance to do this.
Firms that engage in high-frequency trading include proprietary trading desks at a small number of major investment banks (like Goldman Sachs and Merrill Lynch), hundreds of the most secretive proprietary trading groups (like Wolverine, IMC and Getco) and less than 100 of the most sophisticated hedge funds. As a rule they tend to be secretive, stealthy, smart and relatively unknown.
High-frequency trading strategies are highly dependent on ultra-low latency. To realize any real benefit from implementing these strategies, a firm must have a real-time, colocated, high-frequency trading platform where data is collected, and orders are created, routed and executied in sub-millisecond times.
Since many high-frequency trading strategies require transactions in more than one asset class and across multiple exchanges, appropriate infrastructure is required to facilitate long-haul connectivity between different data centers.
The competitive advantage of a high-frequency trading strategy dilutes over time. Although a firm's high-level trading strategy may remain consistent over time, its micro-level strategies are constantly altered for two important reasons. Firstly, because high-frequency trading depends on extremely precise market interactions and security correlations, traders need to regularly adjust code to reflect subtle changes in the dynamic market. Secondly, competitive intelligence is so good across rival trading firms that each is exposed to the increasing susceptibility of their strategies being reverse-engineered, turning their most profitable ideas into their most risky.
On 24 July 2009, Karl Denninger