TOKYO -- In the age of ultra-high-frequency trading, financial institutions are turning to artificial intelligence to improve their stock trading performance and boost profit.
One such company is Japan's leading brokerage house Nomura Securities. The company has been pursuing one goal: to simulate the insights of experienced stock traders with the help of computers. After years of research, Nomura is set to introduce a new stock trading system for institutional investors in May.
The new system stores vast amounts of price and trading data in its computer. By tapping into this reservoir of information, it will make assessments -- for example, it may determine that current market conditions are similar to a moment two weeks ago -- and predict how share prices will be trending a few minutes down the line.
Such computer-assisted predictions may prove to be wrong. But Nomura is employing AI so that the new system will learn more on its own and enhance its price-prediction ability as it gains more data and experience. "It has been about two years since we started working on this research in earnest," said Satoshi Kashihara, head of Electronic Trading Services at Nomura Securities. "Our internal research shows that we have made progress."
In a similar move, Goldman Sachs Asset Management of the U.S. has developed an AI-based system that can analyze as many as one million analyst reports a year and identify any factors affecting share prices. If the system comes across a number of negative expressions in those reports, such as deteriorating earnings, it will automatically revise down investment assessments.
"We use computers to discover bits and pieces of information humans tend to overlook and use such data for making investments," said Masahiro Uchiyama, head of the quantitative investment strategies team at Goldman Sachs Asset Management (Japan).
Rise of AI-assisted trading
These moves are part of a broader shift toward machine-assisted stock trading throughout global markets.
It is essential to have machines in place in order to get ahead of machine-controlled markets, said Nicholas Carrigan, Senior Deputy Head of Equity at SMBC Nikko Securities, who has led the development of a new trading system at the brokerage. Introduced last year, the system automatically buys stocks when it assesses that stock prices have fallen below reasonable price levels.
As financial institutions rush to adopt AI to beat the market, some turn to Kiyoshi Izumi, professor at the University of Tokyo, who specializes in financial data mining and artificial market simulation. Izumi is a pioneer in AI research in Japan who has started joint research projects with five financial companies, including Nomura Securities.
Ultra-high-frequency trading has ushered in the age of this machine-controlled investment trading. In March 2014, U.S. investment company Virtu Financial announced results of its computer trading over 1,238 days, stunning many market participants around the world. The company said it had made a profit on 1,237 days and posted a loss only on one day.
Virtu Financial has been going strong since then and is now focused on the Japanese market as well. Japan is a highly profitable market for Virtu, CEO Douglas Cifu said last month. In Japan, the U.S. company mainly uses a method that involves placing both sell and buy orders simultaneously. For instance, it places sell orders at 100 yen ($0.9) and buy orders at 99 yen at the same time.
It normally takes time for trading to be completed, and prices also fluctuate accordingly. However, if Virtu Financial gets ahead of all other orders by using high-frequency trading, it can stand to gain a profit margin of 1 yen. The company aims to earn profits little by little, just as Japanese Major League baseball player Ichiro Suzuki earns his hits, said Cifu.
On the Tokyo Stock Exchange, high-frequency trading via the TSE Co-location service accounted for 40% of total trades executed last year. As its name indicates, the Co-location service is part of the TSE's connectivity services and is located between the trading systems and the market information systems to increase transaction speeds.
In the U.S., meanwhile, the government is moving to tighten control over high-frequency trading. "Even so, moves to accelerate the speed of trading and further automate stock trading are irreversible," said Izumi of the University of Tokyo.