Citi's New Algorithm Draws On Artifical Intelligence
June 1, 2009
Citi has introduced a second-generation algorithm for executing trades in foreign currencies that draws on sophisticated artificial intelligence to minimize market impact and reduce portfolio transaction costs for hedge funds and other institutional investors.
The new execution algorithm, called Silent Partner, knows what time of the day on any given day how much liquidity will be consumed and how much will be left. The model compares real-time data to historic data on the expected performance of different currency pairs in similar circumstances, depending on the time of day and available buyers and sellers. The algorithm can then move slices of the trade, to minimize impact and costs, and achieve liquidity at the best times.
"We have put a lot of intellectual property into the design of Silent Partner making it a sophisticated execution algorithm," says Andrew Coyne, global head of eCommerce and FX prime brokerage at Citi.
Executing large trades in the foreign exchange market requires firms to identify the presence of available chunks of currencies, retaining the anonymity of buyers and sellers.
"It isn't about latency but about executing with a light footprint," says Jeff Feig, global head of G10 Foreign exchange at Citi. "With larger trading amounts, the first-generation methods of execution don't reduce transaction costs because the market will move away from the client.''
Feig describes Silent Partner as a passive algorithm, which figures out what percentage of market liquidity to consume to get an order done as quickly as possible so the rest of the market won't notice.
Silent Partner works with all the major currency pairs and the most liquid emerging market currencies such as Singapore dollars, South African rand, Turkish lira and the Mexican peso. The algorithm will execute FX deals across Citi's internal liquidity pool as well as EBS, Reuters RTFX, Hotspot FX, Currenex, Lava FX and the Chicago Mercantile Exchange.
Citi's customers can access Silent Partner through Thomson Reuters or Bloomberg's FX order management screens or phone orders through Citi's sales desks. Clients can define the term or the base currency and the maximum and minimum level they want to trade; they can also limit the start and end times to be executed. If customers do not set parameters, Silent Partner will determine the right time to complete the order, on its own.
Citi's launch of Silent Partner reflects the rapid growth of the FX market and increased use of algorithms thanks to the emergence of electronic communication networks which attract buy-side interest. Aite Group, a Boston-based research firm predicts that algo trading will increase from 7 percent of global FX trading in 2006 to about 15 percent in 2010.
While some large hedge funds developed their own algorithms, leading FX banks such as Citi have found plenty of opportunities in providing market aggregation, creating sophisticated order types, and implementing smart order routing technology for institutional investors. By complementing human spot traders, the banks can improve productivity and profitability.
The benefits
Foreign exchange remained one of the few sources of steady revenues for global banks last year, which charge up to $3 per $1 million of currency exchanged, for access to generic algorithms and as much as $7 per $1 million to use more sophisticated algorithms. Those fees are separate from charges to access electronic pools of liquidity.
