|An Algorithmic Explosion|
Trading electronically to a specific benchmark is sweeping the financial markets. But is it delivering on its promise?
Investment firms may not always have a herd mentality, but when it comes to algorithmic trading, there's a veritable stampede.
Take Citigroup. About a year ago, Citigroup's Global Corporate and Investment Bank began making major changes to better position itself for the algorithmic trading movement. These changes started with a joining together of six businesses, involving: alternative ways of trading traditional single stock securities; program trading; connectivity; soft dollars-direct market access; algorithmic trading servers; and global transaction cost analytics, says Will Geyer, managing director and global head of alternative execution. The restructuring also involved setting up regional hubs in the US, Europe and Asia.
Then Citigroup began looking toward acquisitions. The first, Best Execution Comparison Services (BECS) a global, Internet-delivered, post-trade reporting platform, was a seemingly small play that would turn out to be a keystone in the algorithmic trade offering. Then in July, James Forese, the firm's head of global equities, announced a major acquisition in the direct market access space. The firm had decided there were well-developed solutions in the market. With Lava Trading being a leader among them, Citi bought the company earlier this year.
If this strategy sounds revolutionary, it's time for a quick reality check: Many of the big firms on the Street already did or are planning to do the same thing.
Algorithmic trading, often confused with a range of trading types and styles, is, in the simplest terms, electronic trading to a specific benchmark. While it could be used for program trading, and some of the algorithms hail from the program trading space, it is often used for trading single stocks, says TowerGroup analyst Gavin Little-Gill.
And most importantly, it's taking off like wild fire.
A confluence of factors is causing this inferno. On the market forces side, just as ECNs splintered the market, decimal pricing diminished the volumes of stocks that had been available at prices of fractions of a dollar into smaller pools available at prices that differ by just a penny. This fragmentation has pushed traders to find more creative ways to move orders.
Meanwhile, one of the more obvious benefits is the nature of the beast itself: measurability. When everything is pegged to a benchmark, by definition you have a measure of how you're doing.
"At the end of the day it facilitates back-end reporting," says Little-Gill. "After the fact, it's showing what's happening and how the performance went."
There are two main themes that run through most of the algorithmic trading efforts currently on the Street. The first, as highlighted by Citi's Lava acquisition, is direct access.
Brian Fagen, managing director in the institutional equities division at Morgan Stanley, sees this trend as a direct result of the rise of ECNs and of market fragmentation that drove traders to use electronic tools to access the market in different ways. more liquidity became represented in that way, quant fund traders began to be a bigger piece of liquidity. That drove more to want access in a different way. Some might say that anywhere between 40 and 60 percent of order flow is controlled electronically," he says.
The other trend is pre- and post-trade analytics. As there are many newcomers from the buy side to the algorithmic trading space, it seems many firms are trying to make the entrée easier with analytic tools that will help clients decide which algorithms they should use.
"What we are trying to achieve is integration," says Dave Cushing, managing director of program trading analytics at Lehman Brothers. "The first step was to make pre-trade analytics tools readily and easily available on the desktop. The next step is to tie them more tightly to an execution environment so that a client can obtain analysis relevant to the context in which they made that request."
Citigroup plans to leverage Lava's technology for its algorithmic trading servers. "The unique order types and network effect products will help us improve the performance of our algorithmic trading servers," says Citi's Geyer. As Lava has traditionally been a vehicle for sell-side institutions, the firm plans to continue that model while encouraging usage among its clients and those of other participant firms.
"Citigroup will just be offered as another destination within Lava, either as an API or through the front-end GUI," says Geyer.
The firm's algorithmic trading offering will focus on four main algorithms that can be customized by clients: implementation shortfall or arrival price; market on close (MOC); intra-day VWAP and a fourth "opportunistic" model best suited for small-cap stocks that might not have to be traded in a day. This fourth method has a size-of-trade sensitivity analysis that regulates its market impact by passively executing against contra side flow.
The BECS acquisition comes into play in tying the firm's algorithmic servers to pre- and post-trade analytics.
"BECS captures a fund's execution data, which allows us to tailor impact forecasts based on their historic experience. Those predictive statistics are then going to be fed into our servers as adaptive parameters," says Geyer.
And in terms of delivering access to the algorithms, the Lava platform is key. "That was a key enabler for our global strategy to share and develop algorithmic trading intellectual property between our three hubs," says Geyer.
Early to the Game: Goldman Sachs, Morgan Stanley
Goldman Sachs can claim a history of emphasizing algorithmic trading and direct electronic access. It can point to the Hull Trading acquisition in July of 1999 as an early step along that road. In announcing the acquisition, the firm called it part of a strategy of expanding electronic market-making capabilities. The acquisition the following year of Spear, Leeds, and Kellogg brought in the REDIPlus direct access system that became a key component of the firm's direct access infrastructure.
REDIPlus now offers a suite of sophisticated order types, access to Goldman Sachs Algorithmic Trading (GSAT), the firm's proprietary algorithmic trading tools, and for the past six months, it has offered the Guide, the firm's portal to algorithmic order entry and pre-trade analytics.
The GSAT system's four main algorithms are VWAP, Percentage of Volume, Piccolo and 4CAST, an algorithm that takes in data such as real-time market depth during the day to help forecast the impact of a trade, says Andy Silverman, US head of GSAT. A fifth algorithm, Navigator, was launched in version 4.4 of REDIPlus and expanded functionality is expected soon.
"It's the first foray into an algorithm of algorithms. A user will be able to take a list, send it into Navigator, and Navigator will identify the appropriate algorithm to send it to," Silverman says.
In the future, there will also be a possibility to switch algorithms mid-stream if market conditions change, or if at a certain point the smaller amount left to trade makes it more appropriate for a different algorithm, Silverman says.
The pre-trade analytics of the Guide include graphical tools that offer a visual description of what an algorithm will do, explains Greg Tusar, head of development for REDIPlus.
"When you enter a proposed trade, it will give you a visual representation of the impact cost and the volatility," he says, commenting on the potential for adverse price movement. "Access to the Guide has been one of the most important enhancements we have made this year."
The firm is working on a small-cap algorithm, and, separately, it's studying index impact. "In other words, how do I trade Amgen versus a bio-tech index?" says Jana Hale, global head of GSAT.
The system also includes integration with post-trade reports that can guide clients with quickly available data on how efficiently trades have been executed. REDIPlus has similar algorithms for execution in other asset classes, including futures time slicing (TWAP) and options smart routing.
"A key to our strategy is that we work closely with other products within the firm. What we are trying to do is look at a client holistically," Hale says.
Meanwhile, Morgan Stanley developed Benchmark Execution Strategies (BXS) in 1996 as a tool for its portfolio-trading desk and has since integrated the technology across its equity division, including cash equity, options and futures. BXS, which became available to clients in 2001, is accessible through Morgan Stanley's client portal, Passport, through a direct FIX connection or via order management systems (OMS) such as Bloomberg and Macgregor. The technology includes four core algorithms: VWAP, Arrival Price, Close, and Target Percentage of Volume.
"We built strategies based upon benchmarks because we believe traders, like portfolio managers, are benchmarked for performance," says Morgan Stanley's Fagen. "The manner in which a trade is executed determines if optimal performance is achieved. Once a trader understands the objective of a strategy, he or she can select the benchmark that delivers the best results."
In using the strategies internally, one of the areas that Morgan Stanley found further tools would be useful was in measuring P&L.
"Measuring performance is crucial, whether for risk P&L or implementation shortfall," says Fagen. "Measuring P&L is easy, but it gets difficult when you get into complex order types. It is important to have a tool that helps calculate impact."
Morgan Stanley offers clients tools to conduct pre- and post-trade analysis. BXS Navigator, its pre-trade analytical tool, estimates the impact and risk of a particular strategy while Execution Performance Attribution (EPA) measures the performance of the strategy against multiple benchmarks.
In addition to providing the technological applications, Morgan Stanley helps clients use them. "Part of our service is a consultative one. We work with clients to help them understand their order flow and trading performance and look for ways to improve costs so they can effectively utilize our technology," Fagen says.
Turning Things Around
In June, Bank of America Securities' Electronic Trading Services (ETS) group unveiled a Premier Block Trading (PBD) system that "reverses the flow of the algorithm" says Rob Flatley, managing director of ETS. In other words, rather than simply delivering access to the market, the system commits Bank of America's capital to execute trades.
"We provide a real-time, two-way offer and for the cost of our premium up front, traders don't have to risk market impact. We can use our algorithmic strategies to unwind rather than sending to an algorithm up front," Flatley says.
The ETS group, which Flatley runs, debuted at the start of the year. Two key acquisitions—Vector Partners LP, a quantitative broker dealer specializing in portfolio trading strategies in January 2003 and Dallas-based Direct Access Financial Corp. (DAFC), a provider of direct access trading technology in February—have provided much of the technological foundation for the group. Those two pieces brought in the algorithms and the DMA technology, respectively.
"We spent hundreds of millions of dollars to get into this space," Flatley says.
Lehman: Partnering with OMSes
Lehman Brothers' Lehman Model Execution (LMX) system has been used internally by traders and program traders for years, but recently, the firm has made a spate of announcements about LMX becoming available directly through the likes of Bloomberg, Neovest and Macgregor. And there's more to come. "Watch this space," says Jeff Wecker, the firm's global head of electronic client services.
The firm began the process of creating a server-based environment for its algorithmic system two years ago, Wecker says. By adjusting the system to fit a more standardized server architecture, it became easier to test and deliver connectivity to customers. The firm has rolled out the more commonly used algorithms through the system, which allows them to all be customizable.
Dave Cushing, managing director of program trading analytics, is responsible for the rollout of client-facing algorithms and analytics, including transaction cost measurement tools and pre-trade analytics.
The company's Web-based pre-trade analytic engine, Portfolio WebBench, is being made available to be accessed from OMSes much like LMX is accessed today. As its name indicates, Portfolio WebBench originated in the program trading space, but the system has been reworked so that it now offers analytics for single stock trading as well. This latest version has been released internally and will be made available to external clients by the end of the year.
"Whether it's a single stock or a program trade, the analytic information provides recommendations around execution strategy and makes it easy to act on those recommendations," Cushing says.
Next year's theme will be an emphasis on post-trade analytics.
"That means that at any point during or after a trade, a client can get a series of diagnostics and either act on them or learn for the next time," Cushing says.
So for example, if a client was working on a large trade and halfway through saw that it was incurring excess market impact, it could change strategies and slow down the trading.
Bank of NY Turns to Sonic
In March, The Bank of New York (BNY) purchased direct access firm Sonic Financial Technologies, LLC. As a result, through either direct access or broker-assisted trading, clients can use the BNY Brokerage standard suite of algorithms as well as algorithms that the firm customizes for clients.
Because BNY Brokerage had been using Sonic as a vendor for a year prior to the acquisition, the two companies' systems were already integrated, and extending the platform out to the customer was relatively simple, says Candace Cline, managing director and head of Direct Execution Services (DEx). The next version of the technology that the firm is working on will allow clients to deliver orders directly to brokers on the floor of the New York Stock Exchange.
The firm offers a set of the most popular algorithms, as well as a multi-day VWAP that was built at the request of customers. It also does extensive customization of algorithms for clients, though Cline says she doubts these will eventually evolve into canned algorithms.
"Many clients don't want anyone to see what they are doing because they are fearful of reverse engineering or leakage. That's the reason they work with an agency broker. They are that guarded because it's how they make their money," she says.
Meanwhile, the firm's algorithms are becoming increasingly granular in terms of the market data they track and use. "There is an increased focus on the logic surrounding the placement of orders into the marketplace—for example, specific rules determining where to send orders, when to send orders, what the order size is and at what price. These rules take into consideration not only historical trading data, but also current market conditions," says Cline.
Derek Morris, senior vice president for program trading sales at BNY Brokerage, says, "With all the market data that gets generated, the point is to try to do what a human eye can't do in terms of filtering noise. The process for a human trader is to take in these inputs and try to make sense of them to try to decide when and where to execute. The goal is to use algorithms to assist the trader in determining how the stock is trading, whether it's gaining momentum or reverting towards some mean price. We use sector performance as another input in the process."
Icing on the Cake
Finally, the ultimate sign of algorithmic trading's popularity is a new independent portal that was born this fall. In October, institutional brokerage and technology firm Electronic Specialists (also known as ESP) introduced an algorithmic trading portal offering clients anonymous access to 30 of the most used algorithms. The portal, called Electronic Algorithm Routing Network, or EARN, counts Credit Suisse First Boston and BNP Paribas among the early participants.
Though other firms have multi-broker portals, the fact that this is independent and brand new harkens back to the start-up fever in the early ECN days.
The idea for the portal came from customers of the company's institutional brokerage business, says senior managing director David Share.
"There's such a proliferation of algorithmic offerings that integration becomes cumbersome because you have to have direct relationships. We decided to create one-to-one relationships with each of these desks," says senior managing director Scott Forbes.
by Renee Wijnen Caruthers