The rise of the automated trading desk

artificial intelligence meets HFT

At the trading desk of the future, humans may not be at the table.

Automated trading is a big piece of today’s market volume. According to market research firm Preqin, 1360 hedge funds – or about 9% of all funds – trade primarily with help from computer models. These quantitative funds manage about $197 billion in total. Nor are hedge funds the only ones using high-frequency trading (HFT) tools and techniques — many prop desks are doing it, too.

From the perspective of buy-side and sell-side institutions, HFT volumes are meaningful. During 2009-2010, HFT accounted for nearly 70% of all market volume. Experts believe that number has dipped in recent years, but competing in today’s market — as a trader or investor — means identifying the counterparty on the bid- or ask-side of the trade. It means understanding what the market environment will look like a second after you set a limit order, so that you get the best pricing possible.

And to do that, technology is increasingly stepping in.

When much of the market volume is controlled by algorithms, human traders must determine which algorithm to use at a given time. Some funds are employing automated technology — called an algorithm switching engine — that acts as a mechanism to control an algorithm’s parameters based on market microstructure variables (like order flow). Today’s trading has evolved from merely building algorithmic strategies to optimizing the choice of the right algorithm at the right time.

“It turns out order flow is predictable, so there’s predictability in algorithm performance,” commented Henri Waelbroeck, director of research at Portware, a developer of Execution Management Systems. “With the right tools, you can anticipate performance.”

Waelbroeck, who previously was a professor at the Institute of Nuclear Sciences at the National University of Mexico, is Portware’s director of research. He describes two ways that Portware clients benefit from implementing algorithm switching technology. The first is that their participation rate improves. There’s a certain level of adverse selection in entering a trade into a market controlled by other algorithms — you get filled faster when the market is coming towards you than when it’s going away from you. Algorithm switching enables an automated response to market conditions. Second, cost quality should improve because trading strategies should get better prints. This level of automation should provide tangible, demonstrable results from investors deploying such tools.

Alpha profiling is another way automation is taking over the trading desk. Developed in 2010 by Portware, this artificial intelligence tries to understand when a portfolio manager’s order is similar to other orders out in the market. If it is, there’s an underlying urgency to get the order done but if the order is contrarian, the technology can execute as slowly as possible to get the benefits of price improvement.

Waelbroeck believes many of the existing risk management solutions on the market today are too often predicated on correlation and volatility of the recent past when in fact, risk is all about what can happen in future. Risk scenarios, by design, must look at factors that can dramatically change in the future. When discussing Portware’s AI,Waelbroeck provides an analogy to human intelligence: To a large extent, as humans, we analyze data not for what is says about the present but what it says about the near future. Few machines are designed this way.

Systems like Portware’s have introduced artificial intelligence into the trading desk to analyze and act upon big data. Similar things are happening inside banks to the equity analyst role as the use of technology is increasingly central in equity research.

In the fourth quarter of 2015, FactSet announced it would acquire Portware for $265 million, making it the largest acquisition in its history. The idea for the tie up seems to be that FactSet, one of the largest providers of financial information and analytics, could provide Portware with the data and analytical tools to take its Execution Management System to another level.

“HFT isn’t a zero sum game,” said Portware’s Waelbroeck. “But the games are becoming more difficult and an arms race is underway as a deterrent to some of the explosive techniques that people used to use. Institutions have gotten smarter and can protect their business interests against people trying to profit from order flow.”