Algorithmic trading is growing fast in the U.S. and European markets, where its volume exceeded those of traditional traders.

It’s an inevitable evolution, which has flourished with the technology bubble of the late 2000s, and which does not seem to slow down according to the statistics. The last three years, the percentage of automated trading volume over the total traded volume has increased by a third on the London Stock Exchange, and is now around 65%, not far from the New York Stock Exchange . In the U.S, the automated trading only accounted for a third of the market in 2005. It has doubled in four years. It is still difficult to distinguish a decision based exclusively on computer analysis versus a trader decision drived mostly by systematic scans.

According to John Bates, managing director of Progress Apama, a company providing computer-based models for financial analysis , the "algo trading has been for years an invaluable tool for financial services. It is now almost impossible for hedge funds traders, pension funds, mutual funds or other institutional players to do their job effectively without this tool. Traders are becoming more confident with algorithmic models for a better view on various liquidity pools "

In some markets, algorithmic models account for 80% of trading decisions, and even the most reluctant users join anyway: the foreign exchange market, which stood at 25% in 2006, is catching up. The future markets are also catching up the trend with a forecast of 20% of total volume automatically handled by computer in 2010. Bond markets will also follow.

The advantages are related to the speed of decision-making - even faster than classic trader - and more generally to costs reduction in order to attract more customers.

One category of algorithmic trading is the "news trading" which relies on real-time automated trading signals - buy / sell - based on a detailed analysis of news related to the companies results or activities. The spectrum of instantaneous information analysis is wide, with a growing complexity.

Words and sentences appearing in the news wires are treated to indicate the "sentiment", either positive or negative, with a relative impact on the final decision. Each type of information has clearly a predetermined score. The purpose is to provide computer processing systems with a maximum "human" sensitivity, while streamlining as much as possible that treatment with an increasingly, but still relevant, high volume of news . The information covered ranges from corporate profits or losses, climatic or political events in a specific geographic area, including management changes, rumors of a merger, economic structural statistics...

More and more companies require non-financial news to be incorporated into algorithm. That is what has led Thomson Reuters,the financial news agency, to create a new wire that can be directly integrated to algorithms or computer models in order to forecast volatility and enhance decision-making process or risk management.

Thomson Reuters has also launched a series of indices updated in real time to indicate an unusually large volume of news in given sectors, and thus alert the "news strading" systems.

Algorithmic models are expected to become more complex, both because the broadcasted information volume is getting higher and higher, and because competition leads firms to develop more efficient analytical systems. Their ability to design high-end models depend on their profits and competitive advantages

However, the "news trading" has limitations. According to Sang Lee, a partner at consultant Aite Group, "we find in algorithmic trading field that most players want to handle more and more data. But this amounts to ignoring the fact that the data used by the algorithms are often redundant or relatively dated"

Even backdated ... In September 2008, southflorida.com, an American cultural information website, has mistakenly published a 6-years old news on the bankruptcy of United Airlines, which were then resumed on bloomberg website. "News trading" softwares operated faster than sound, leading within minutes the share price to $3 from $12.5 at market opening!

Algorithmic models are likely to remain much longer in a search of a total infallibility.