In the race to get their algorithms online and accessible to institutional customers, many brokers are eager to put their logos on the desktops of order-management systems (OMS).
"Obviously, we always have a queue of brokers banging on our door to do more business," says Eric Bernstein, vice president, product management, at Linedata Services, which owns Longview Trading. So far, Longview has built about a dozen interfaces to leading brokers such as Banc of America Securities, Morgan Stanley, Goldman Sachs and JPMorgan Securities.
Brokers supply the vendors with specifications to create customized FIX tags, based on the Financial Information Exchange (FIX) protocol. The vendors use the FIX tags to create order tickets. "We make sure we're reading the tags properly and passing the orders to that destination," says Bernstein.
But the pace at which the OMS vendors are able to create these interfaces is not as fast as some brokers would like. "Brokers are asking why we can't do this quicker," says a developer at one of the leading vendors that's building algorithmic trading interfaces.
The conflict inherent in algorithmic trading is that brokers need to create interfaces to front-end trading systems and order-management systems used by their buy-side customers, but sell-side brokers do not want to disclose the secret details of their proprietary algorithms, says Scott Atwell, manager of FIX trading and connectivity at American Century Investments. Atwell is co-chair of industry group FIX Protocol Limited's (FPL) Global Steering Committee and Global Technical Committee. Because the algorithmic trading business is so new, he continues, brokers are trying to create a better mousetrap, and they don't want to tip their hands.
Can FIX Work?
While everyone agrees that the strength of FIX has been its flexibility in supporting innovation, the lack of a standard way for coding the strategies into the message can create problems. "FIX in general is what I call an interpretive standard. They have tags and fields for structuring messages, but people interpret it differently," says Rob Flatley, managing director, electronic trading services, at Banc of America Securities. "Because there is no standard, the variance from broker to broker can be quite varied," he says.
If a buy-side customer has six algorithmic brokers, and the specs for each broker's FIX tags are different, but the client generally wants the ticket windows to look the same, "The vendor has been put in a position to normalize the FIX messages that deviate from broker to broker," says Flatley.
Another obstacle is that about 90 percent of the trading community is using FIX 4.0 and FIX 4.2, and these earlier versions of the protocol don't support algorithmic trading. The latest version, FIX 4.4, however, contains fields called TargetStrategy tags that address algorithmic trading.
Most firms, though, have not yet upgraded to this version of the protocol. So, all the brokers have told the vendors to use user-defined fields or custom tags to name the strategy and structure the order ticket. This will enable buy-side customers to fill in the parameters of a particular strategy, such as start time and end time, and route the order to the correct broker's engine.
But, because the brokers define the custom FIX tags differently, this is leading to a lack of coordination. "Everyone is using FIX to do algorithmic trading," says American Century's Atwell. "The challenge is, when you use one broker's algorithms, they tell you, 'I want you to do it this way.'"
Atwell believes that the trading community can use FIX as a framework for creating standard interfaces to brokerage firms' algorithmic trading servers - at least for the most common algorithms - without giving away their secret sauce. "Although the internal workings of algorithmic engines really are proprietary, there really is an opportunity to ensure there is consistency around that so that we don't make [the interfaces] complicated and difficult to use," Atwell told attendees of FPL's November Americas conference in New York City.
FPL Addresses Algo Trading
In August, FIX Protocol Limited created the Algorithmic Trading Working Group to help the industry solve the interface problem - regardless of what versions of the protocol firms are using today - and to try to find a flexible framework for the community to use in the future, says Atwell. Eighty-four members have signed up to participate in the discussions; the group is chaired by Chris Sims, head of information systems development at Gartmore Investment Management in London.
"The goal of this working group is, No. 1, to get a community of people who are using algorithms, who are interested in this, together to try to inventory how algorithms are being used," Atwell relates. The committee has already identified four ways that firms are developing FIX interfaces to algorithms. One method is when firms like CSFB register tags on the FIX protocol Web site - other firms can then use the tags CSFB has already registered. Another option is applying the FIX 4.4 tags to earlier versions of the protocol.
Mark Chazen, director of electronic trading and connectivity at Citigroup's alternative execution group, says the purpose of the working group is to create a standard going forward, not to define what the algorithm is going to do or figure out how the quants are designing the mathematical models behind the scenes. The focus is on "how to make it easier for vendors, brokers and investment managers to create the interface to strategies via an OMS," he says.
Jeromee Johnson, vice president of professional services at institutional brokerage firm UNX, thinks industry consensus is building around how to label five or six trading parameters, such as start time, end time, the handling of different time zones or how to handle aggressiveness. "If five firms are doing start-time, end-time and participation-rate in five different tags and different places in the message, that's an implementation challenge," the connectivity expert says.
Close to the Vest
Today, there is a reluctance by the brokerage community to share its algorithms that is making it more difficult to create the interfaces. "On the buy side, we'd really like the brokers to open up a bit in terms of what these interfaces are [doing]," says Atwell, whose firm builds its own interfaces to broker algorithms because American Century has a proprietary OMS. "The algorithms themselves may be rocket science, but the interfaces are not," Atwell adds.
On top of that, Atwell says a number of brokers are paying for OMS projects to develop the interfaces to their algorithmic engines so that they get priority over competitors. Compensating the OMS vendors supports the revenues of vendors, and wins the brokers shelf-space, but it may not be the best thing for the industry, says Atwell, who notes that the practice could further fragment the protocol.
Still, Banc of America Securities' Flatley says it's not unethical for the OMS vendors to make money. "These guys are in business; they're not charities," says Flatley, who adds that his firm has no problem paying the OMS vendors if they provide good service. Some are looking for a one-time payment for development and maintenance, while others want transaction revenue per share or an ongoing routing and connectivity surcharge.
Flatley says it can take as much as six months to integrate a broker's FIX tags into an OMS - factoring in two to three months to build the tags, then testing and certification with each client's FIX order routing network, and waiting for lawyers to create the documentation. If any of these parties drops the ball, development of the interface can take even longer.
No Stopping Progress
Despite the bottlenecks, OMS vendors are cranking out the interfaces. "We're trying to find a way to get new FIX interfaces into our product and out to customers," said Tom Driscoll, vice president at Charles River Development, speaking at the FPL America's conference. "We're down to days," Driscoll said, noting that the OMS vendor has 175 buy-side clients, of which 90 are involved in equity trading.
In the last year, Charles River has added four algorithmic venues, and Driscoll related that there's a large number coming up in the next quarter. Since brokers upgrade their strategies often, Charles River is using XML messaging as a technique to keep up with the new algorithms. When asked at the conference how often his firm has to work outside the standard FIX message, Driscoll said, "For us, it's constant; we're constantly re-mapping tags, but it's not programming."
Linedata's Bernstein says that FIX tags for algorithmic strategies like Volume Weighted Average Price and Time Weighted Average Price are becoming "pretty standardized." But with more complex algorithms, with which the user can enter trading parameters around volatility, there is going to be more customization. Clearly, it's in the interest of OMS vendors and the brokers to standardize as much as possible if they want to get their algorithms out to customers. "The more standardized the [FIX] tags become, the easier it's going to become to support [algorithmic trading], and the easier to develop [interfaces]," says Bernstein.
Connecting Via FIX To Algo Engines
The FPL Algorithmic Trading Working Group has identified four approaches for connecting to algorithmic trading engines. The common options are:
1. The "Text Option" - Using pre-defined text in Target SubId (57) or Text (58) to identify the desired strategy.
2. The "18/40 option" - Using ExecInstruction (18) or OrdType (40) to identify the desired strategy.
3. The "6000 User Defined Option" - Use of the custom tags (6061-6067) registered by CSFB.
4. The "4.4 TargetStrategy option" - Using the Target Strategy tags (847-850) available in Fix 4.4 even in pre-4.4 versions.
Source: FPL Americas Conference, Nov. 17, 2004
For more information, visit www.fixprotocol.org. Ivy is Editor-at-Large for Advanced Trading and Wall Street & Technology. Ivy is responsible for writing in-depth feature articles, daily blogs and news articles with a focus on automated trading in the capital markets. As an industry expert, Ivy has reported on a myriad ... View Full Bio