Mizuho Financial Group Inc. will start artificial-intelligence trading to bolster its Japanese equity business, according to people with knowledge of the plan.
Japan’s third-biggest lender will begin offering an algorithm-based AI trading service to some large institutional clients in Japan and elsewhere in Asia, the people said, asking not to be identified because the plan is private. The product attempts to predict how stock prices in Japan will change over an hour and find the best time to trade, they said.
Global banks are seeking to improve the quality of their trading execution and equity research before the European Union’s revision of the Markets in Financial Instruments Directive, or MiFID II, comes into force in January. Under the new rules, brokerages will be required to separate research fees from trading commissions to ensure investors pay fair prices for the services, a move that may spur competition.
Mizuho’s brokerage unit will initially target 500 stocks on Japan’s benchmark Topix, and may increase coverage to 1,000 companies including firms listed on the Tokyo Stock Exchange Mothers market, according to the people. It plans to roll out the service to all of Mizuho’s trading execution clients, which number in the hundreds, by the end of the year, the people said.
The AI platform uses about 5,000 data points, including order-book information, historical prices and news sentiment for the targeted companies, as well as exchange rates and commodity price changes, to find anomalies and patterns relating to a stock’s movements, the people said. It will enable clients to trade large volumes of stocks at an optimal time, lowering costs and reducing market impact, they said.
The brokerage tested the platform from May to August, making 500 transactions each month valued at around ¥30 billion ($270 million). Trading prices were 1 to 1.7 basis points more favorable than the volume-weighted average price, according to the people.
Hiroki Saigusa, a spokesman for Mizuho Securities Co. in Tokyo, declined comment. Bloomberg LP, the parent of Bloomberg News, provides execution services for firms complying with MiFID II requirements.
Unlike regular algorithmic trading, where transactions are carried out swiftly according to set instructions, AI trading involves learning by the machine to improve its predictions. Mizuho Securities formed an AI team in 2015 that consists of about 10 specialists including doctorate-level computer scientists and professionals in algorithms, the people said.
MiFID II requires buy-side clients to take “sufficient” steps to achieve the best execution or price. This means institutional customers are increasingly looking at more scientific methods to analyze prices across a range of liquidity providers.
Instinet, the equity execution services arm of Nomura Holdings Inc., is also developing more advanced trading systems in light of the regulation change.
“Under MiFID II, clients are obligated to demonstrate that they are achieving best execution; this has then driven innovation amongst the sell side,” said Patrick Mohr, head of execution consulting at Instinet Pacific Ltd. He said the firm “has already invested in staff and technology and incorporated machine learning into our algorithmic and analytics platform, and seen significant improvement in performance based on this work.”
The impact of MiFID II is expected to be global. In Japan, Daiwa Securities Group Inc., Mitsubishi UFJ Morgan Stanley Securities Co., Citigroup Inc. and BNP Paribas SA are among banks that are taking steps to bolster their equity business before the regulations are implemented.
In a time of both misinformation and too much information, quality journalism is more crucial than ever.
By subscribing, you can help us get the story right.