Computers that read news stories and use the information gleaned as a basis for trading will become widespread over the next five years, Reuters said yesterday.

The information provider will make available today a tagged version of its news feed — which provides 8,000 news stories a day — to make them easier for computer-based trading systems to digest. It describes the effort as generating “machine-readable news”.

Peter Moss, the head of Reuters Enterprise Solutions,said: “In three to five years a much more sophisticated set of algorithmic trading will emerge, based on information as well as market data.”

Today about 60 per cent of trading in the United States is conducted by computer; in London the proportion is estimated to be 40 per cent. But the trading patterns used by algorithmic systems are based principally on price and volume data, although the situation is changing fast.

Computers reading news — in effect, trying to simulate human judgment — signals the beginning of the end for the brash City trader. It heralds an era in which human intervention in trading systems will be dominated by programming and monitoring computers.

To help banks and hedge funds to generate their own trading models, Reuters is also making available a news and price archive dating to January 2003. That is intended to help to develop and test trading models based on historic information. Massive amounts of data are involved: the archive will include nearly 12 million news items and pricing information is updated up to 23,000 times a second, which includes data obtained from 258 exhanges and markets around the world.

Richard Brown, of the company’s machine-readable news programme, said that such systems will help computers to gauge how far the market will react to an earnings miss — but may also help to forecast an economic or political surprise by detecting a “flurry of news stories” before the crucial moment.

Initially, Reuters is supplying only its own news stories and it is asking its own journalists to help to tag the stories with five to ten key items of information. Through licensing agreements, the company could supply news stories from newspapers and other outlets.

However, the information provider believes that its clients will develop tagging systems of their own. “This is a stealth technology,” Mr Brown said, noting that many of the company’s clients refuse to disclose how they hope to use the data.

Reuters expects that computers will be able to weight news stories based on the language used to evaluate how tenatative they are — and check the byline to establish whether it has been put together by a specialist in the subject in question.

Times Recommends

• Lies and cover-up at MG Rover exposed by report
• High oil prices push manufacturing costs up
• Primark founder steps down as chief executive