近日，AllianceBernstein Holding LP（后文简称AB）升级了他们的虚拟助手Abbie，后者可以基于定价、交易便利度和风险推荐什么债券最应买卖。
1.0版本只有在明确接受需要购买多少哪一种债券后才能创建订单，而AB首席固定收益运营官Jeff Skoglund表示，2.0版可以找出人们可能忽视的债券而且可以识别人类错误并和其他公司创造的像她一样的机器人交流，同时可以挖掘数据池比如该公司的信用研究库并给出建议。也就是说，Abbie 2.0会检索公司的评级，并从多个角度研究他们的债券比如资本结构、环境、社会和治理风险，然后推荐符合特定基金运作要求的债券。
Greenwich Associates市场结构和科技研究负责人Kevin McPartland去年在该份报告中写道："大数据和机器学习已经被广泛用于增强交易商和投资者的智能。"
The robots have just got a step closer to managing your money.
AllianceBernstein Holding LP upgraded its virtual assistant Abbie so she can now suggest the best bonds to buy and sell based on pricing, ease of trading and risk. Unlike any human, she can scan millions of data points to filter the universe of outstanding bonds in seconds and identify potential trades to portfolio managers using other electronic tools the firm has built.
Abbie 2.0 can identify bonds that people may have missed and will be able to spot human error and communicate with chat bots like herself at other firms, according to Jeff Skoglund, chief operating officer of fixed income. The original electronic assistant AllianceBernstein introduced in January could only build orders for bonds following precise demands.
“Humans and machines will need to work closer than ever to find liquidity, trade faster and handle risks,” Skoglund said. “Our hope is that we grow and use people in ways that are more efficient and better leverage their skills.”
Abbie already supports more than 60% of AllianceBernstein’s fixed-income trades. The upgraded version would target US high-yield and investment-grade bonds initially and expand to other markets in the months ahead, he said. The $550 billion asset manager will still rely on humans to make the final decisions and execute trades.
While the original could only build an order for a bond if she was told exactly which security the portfolio manager wanted and how much, the new version mines data pools, such as the firm’s library of credit research, to make her suggestions. She scans its ratings of companies and their bonds on factors such as capital structure and environmental, social and governance risks and proposes bonds that meet a portfolio’s requirements.
Investment firms are crunching data and harnessing technology to cut trading costs, save time and avoid errors in markets where bonds trade infrequently. They’re also seeking an upper hand in debt dealings as banks pull back from making markets amid post-financial crisis regulations. Investors are concerned that liquidity will matter even more when the years-long bull market in fixed income ends.
Electronic trading is becoming more prevalent in fixed income as banks act as matchmakers rather than holding bonds on their balance sheet and as regulations encourage the shift of trading to exchanges. Over 80% of investors in high-grade bonds use electronic platforms, accounting for 20% of volume, according to a Greenwich Associates report.
“Big data and machine learning are already being applied to augment the intelligence of dealers and investors,” Kevin McPartland, head of research for market structure and technology at Greenwich Associates, wrote in the report last year.
To source bonds that are easy to trade, Abbie accesses another AllianceBernstein system called Automated Liquidity Filtering and Analytics. Alfa gathers bid and offer prices from dealers and electronic trading venues to work out the best ways to trade, amounting to about three million data points a day. Alfa was created in 2015 and sold to Algomi Ltd last year, but AllianceBernstein kept the source code and continues to use it internally. “We expect to be faster to market and capture opportunities we otherwise would not have caught by using this system,” said Skoglund. “There’s a liquidity problem right now that could become significantly more challenging in a risk-off environment.”