FX algorithmic trading is set to grow, but as Frances Maguire asks, will the FX market go a similar route to the equities market, rather than learn from it?
In its recent white paper, entitled 'Algorithmic Trading in the global FX market: The Need For Speed, Transparency And Fairness', FXall outlines how the foreign exchange market is following the trend where hedge funds are successfully applying algorithmic strategies to FX trading to generate alpha, using many of the same methods they have applied in the equities market. According to FXall, the growth of FX algorithmic trading is bringing key issues such as speed, inherent and network latency, transparency and market fairness to the fore.
Accelor, FXall’s FX ECN introduced in February 2007, is one of FX’s newest ECNs. It uses prime brokers as credit intermediaries and provides an anonymous central limit order book that honours price/time priority. Its high-speed matching engine has been architected for minimal latency, with orders acknowledged within 3 milliseconds and filled within 20 milliseconds.
Mark Warms, general manager, Europe, at FXall, says that Accelor uses next-generation ECN technology. “The majority of volume traded over Accelor is handled by algorithmic traders, which have high demands in terms of the performance of the venue over which they trade. Low latency, for example, is a necessity, as the faster trades can be filled or cancelled lowers the risk and the better the algorithmic models can run.”
The equity market has already sorted out many of the issues that the FX market is just facing. FX ECNs are relatively new. A central part of Accelor’s strategy is the decision to operate servers from a central book and a single server, located in New Jersey. Customers can co-locate at that facility with Accelor servers. In addition, Accelor can be accessed through the BT Radianz network, and through a variety of methods, including the internet.
Says Warms: “Where systems exist in multiple locations, customers do not have the ability to route their order to a single server. Problems may arise when you put an order in before someone in another location, yet their order will get matched first. With one location, Accelor can always honour price/time priority”.
Simon Keen, European technology director for StreamBase Systems, which provides banks, hedge funds and exchanges with algorithmic trading solutions, says that the automated trading strategies of the hedge funds and proprietary trading desks are increasingly using algorithms to trade FX. He says: “FX algorithmic trading is mainly in futures contracts and the algorithms are not the same as those used in equities execution, they are more complex. Algorithmic trading, as opposed to algorithmic execution, in all asset classes is seen as substantially more advanced than algorithmic execution and thus us not borrowing much from the latter.
According to Keen, FX algorithmic trading is a very latency dominated market, although the volumes, and thus the throughput requirements, are currently lower than the equities market. The approach to solving this latency requirement is more complex than in US equities. The complexity is brought about by a number of factors, as Keen explains: “The majority of FX trading is performed in London but the venues used to trade on are spread around the world. Hence co-location, which makes sense if you are trading on a single market, is not really applicable. Furthermore, the reality of any trading system is that it consists of many systems from the data feed through to the final trade booking system. The focus is in reducing end-to-end latency, which involves all the systems in the path. And finally because the algorithms themselves can use complex models, these can form a primary part of the latency along with any risk models involved.”
Harrell Smith, co-head of product strategy at Portware, which provides multi-asset automated algorithmic trading solutions, says that the FX and equities markets differ significantly from a structural and regulatory standpoint. In equities, the buyside has the same kind of access, in terms of speed and technology, to execution centres, whereas the FX market still operates as an OTC market, with distinct interdealer and dealer-to-client marketplaces, and has resisted any moves towards an exchange-based trading model.
Smith points out that: “Until very recently dealers did not have much problem executing size and managing market impact in the FX market, but this is changing now that dealer networks are slowly being opened up to the buy-side. This is one of the major issues now facing the FX market with respect to alpha-generating algorithmic trading, and one of the major differences between the two markets from a structural standpoint.”
Historically, the lines between the interdealer and dealer-to-client execution markets were fairly distinct. Recently, however, interdealer platforms have been looking to increase order flow by opening up their systems to the buyside, and the FX market is starting to look increasingly like the equities market, where venues are for the most part open to all market participants.
There are limits to this, but Smith adds that the increased use of electronic trading in the FX means the market is approaching a situation where liquidity is commingled. He says: “I don’t know whether the FX market is learning from the equities market except that many FX dealers are having to learn how to navigate in commingled markets where there is now additional, and very aggressive, liquidity from buy-side firms deploying alpha-seeking FX algorithms.”
There are some major challenges to be overcome by the FX market. Not all FX destinations use the same communications protocol, whereas FIX has become the industry standard for electronic equities trading, there is still some customisation needed in the creation of an FX aggregator, which Portware has developed.
But Mike Thrower, head of sales and marketing at Cognotec, which builds rate management solutions for some of the largest FX liquidity providers, says that in last few years there has been a convergence from a variety of standards in the FX market, akin to the equities market. He says: “The vast majority of our clients have embraced FIX for the FX market and we are seeing a strong drive from our clients, who are looking for ease of integration and rapid time to market, to support FIX, and are increasingly insisting on FIX when hooking up with the buyside. For computer-to-computer conversations to take place efficiently, a common standard is fundamental.”
Furthermore Thrower believes that the ability to trade in large amounts in volatile markets, without moving the market away is attractive to those trading the market for alpha, and that the FX market is too large and too liquid a market for the buyside to impact the price traded.
He adds that even though the FX market is a global OTC market, there are still co-location issues, and where servers are physically located impacts latency. In response to client demand, Cognotec has recently opened a new data centre in Chicago to improve the response times and latency for its FX customers in the US. Both its London and Chicago data centres hooked into the BT Radianz network to enable customers to co-locate.
Dr John Bates, founder and vice president, Apama Products, Progress Software, says that FX has leap-frogged the first generation execution algorithms, such as VWAP and TWAP, that equities adopted. He says: “We have seen increased demand for tools that provide differentiation that ‘black boxes’ cannot provide. With FX algorithms, most have accelerated through the this stage because volume-based VWAP-type black box strategies were not as appropriate to the FX market and because FX algorithms are being implemented as part of cross-asset and multi-leg strategies. Therefore, the only way to effectively implement those strategies was, in effect, to skip past the first generation and move straight to flexible, customisable strategies.”
According to Bates, concerns about latency still exist in the FX market, despite the fact the technology is newer, but that the near myopic focus on execution latency, addressed by strategies like co-location, should not be allowed to obscure the complementary issues that are all part of the trade process, such as risk management. “There is a need for platforms that are flexible in addressing performance requirements and not focused exclusively on delivering orders to the liquidity venue as fast as possible,” he says.
Many of the recent entrants into the FX market have come from the equities market and already have venues that are designed for algorithmic trading in terms of speed and transparency. They are looking for the same requirements in FX and if the FX market can supply them, there is little doubt algorithmic trading in FX will grow to the same heights as equities.
