Inventors of trading algorithms—computer programs that generate buy and sell orders and make lightning-quick trades — have picked up a bit of timeless wisdom from the stock operators of old. “The best of all tipsters,” Edwin Lefèvre wrote in a 1923 fictional account of the markets, “is the tape.”
As the time taken to process computer-generated trades falls to thousandths of a second, algorithms are being created to react to news headlines faster than the eye can scan them. Dow Jones and Reuters, the news providers, now offer electronically “tagged” news products that algorithms pick up to make programmed trading decisions. (Dow Jones claims the business is so secretive that it cannot divulge details of customers.) Britain’s Financial Services Authority, a regulator, also hopes to use algorithms to comb through trading data to find hints of suspicious activity, which it reckons takes place before about a quarter of all takeover announcements.
Algorithmic trading accounts for a third of all share trades in America and the Aite Group, a consultancy, reckons it will make up more than half the share volumes and a fifth of options trades by 2010. On June 18th the London Stock Exchange unveiled an electronic system catering to the growth of algorithmic trading, which cuts trading times down to ten milliseconds. On its first day, it processed up to 1,500 orders a second, compared with 600 using its previous system. The ability to push up volumes should help dissuade customers from moving to faster platforms elsewhere.
The aim is to reduce the delay between order and execution, known as latency. Every moment is crucial in “black-box” and “statistical arbitrage” trading, where computers prowl through the market for price distortions that may last only for a split second. Order-handling algorithms, which break up large trades, must also move faster than the blink of an eye to ensure they get the best electronic prices.
According to TowerGroup, a research firm, $480m is likely to be spent in America this year on developing technology for algorithmic trading. Such is the focus on speed that even location counts. Servers positioned nearest to a trading venue can shave milliseconds of the timing of a trade and get a better price.
Low latency could also help investors get a jump on news of economic data as it flashes across the wires. According to Andrew Silverman of Morgan Stanley the use of news feeds for algorithmic trading is at an early stage. The software, which relies on keywords to generate buy and sell orders, may misunderstand the context surrounding a headline. For example, a market-moving word such as “surprise” may indicate numbers are better, or worse, than expected. Mr Silverman explains that news algorithms are best used with other variables, such as share price and volume, which may reinforce the buy or sell signal.
Now that trading algorithms are reading the news, are they also getting the story faster than journalists can? Regulators suspect that some price movements before takeover announcements stem from algorithms picking up early-warning price signals in the market. Eventually, the news may come from reading the algorithmic trades, not the other way around.