You are here

China’s first AI fund learned from the country’s best traders

[SHANGHAI] China's investors will soon be able to pick machines over humans when trying to profit in the world's wildest major stock market.

Zheshang Fund Management Co, which manages US$6.5 billion, plans to use about 300 investment models to analyse more than 3,000 Chinese stocks for what it says is the nation's first artificial intelligence-based fund run purely on recommendations from its machines.

The "robots" initially learned from the nation's best fund managers and analysts, and after about a year in training are ready to take them on when the fund opens to retail investors next month, according to Zha Xiaolei, head of Zheshang's four-person AI (artificial intelligence) investment team.

While a growing number of hedge funds globally including Bridgewater Associates and Man Group plc have developed AI and machine learning strategies to gain an edge, the results so far have been mixed. Funds that incorporate AI into their decisions have lagged a broader hedge fund index since 2018, according to Eurekahedge. Making headway in a market famous for its unpredictable price swings, where mom-and-pop investors drive 80 per cent of trades, won't be easy.

"Globally AI strategy is at a very early stage, even though it may gain traction in the long run," said Liu Shichen, a Shanghai-based analyst at Z-Ben Advisors, an asset-management research firm. Explaining and selling the concept of an AI fund to a retail audience in China may be a tough task, he said.

Your feedback is important to us

Tell us what you think. Email us at btuserfeedback@sph.com.sg

To build a model that can adapt to China, Zheshang's machines first analysed the strategies of more than 80 of the nation's best-performing fund managers, drawing from public disclosures of their holdings. They were also fed stock recommendations from 500 star analysts and industry experts, and investment ideas from 200 chat groups.

Mr Zha said they took volatility into account from the beginning by targeting Sharpe Ratio, a measure of risk-adjusted performance, as the key indicator for the robots. Diversifying the portfolio also helps contain the volatility, he said.

The machines now analyse 3,000 different data feeds to generate trading signals, and consistent underperformers are eliminated.

"Every robot was assigned one specific goal and they need to reach perfection in doing it," said Mr Zha.

Early results

The fund returned 26.4 per cent between Sept 28 and June 28, according to Mr Zha, a period during which Chinese stocks swung from a bear market into a bull run. The benchmark CSI 300 Index added about 12 per cent.

Mr Zha, the only member in the AI fund team with a finance background, also manages Zheshang's Big Data fund, which outperformed 94 per cent of peers with a 22 per cent return in the past year. Xiang Wei, the co-manager of the AI fund, has a doctorate in computer science and previously oversaw development of the personalised search engine at Baidu Inc.

In a Big Data fund, managers analyse massive volumes of data to make investment decisions, while with machine learning - a subset of AI - computers improve their ability to independently find trading signals as they crunch more data over time.

The biggest challenge, Mr Zha said, is that his machines haven't had sufficient time to learn. The paucity of data on new companies and some industries in China can hinder performance, he said.

Eurekahedge doesn't track any AI hedge funds based out of mainland China, said Mohammad Hassan, its Singapore-based head analyst for hedge fund research and indexation.

AI funds shouldn't be a harder sell to retail investors, Mr Hassan said. Clients will eventually be drawn to performance and other factors such as fund governance, he said. "The AI hedge fund space is still in its nascent stages, but the shift towards more technology, machine-intensive strategies is here to stay."

BLOOMBERG

BT is now on Telegram!

For daily updates on weekdays and specially selected content for the weekend. Subscribe to t.me/BizTimes