In addition to the growth of FX algos deployed on primary CLOBS, the BIS Working Paper Banks notes that, due to the role that such platforms play in the market, algorithmic trading has also had an important impact on price discovery in FX. Furthermore, the paper adds that execution algorithms are now also being used directly by some of the more sophisticated customers in the FX market and that these users “increasingly rely on smart order routing and execution algorithms to spread large orders over time and across multiple electronic venues.” Joel Marsden, eFX Senior Currency Trader at ANZ, says that algo trading literacy is also increasing considerably among the wider institutional and multi-national corporate client base, which is a broadening from earlier investor and asset manager adoption. He explains that with this knowledge comes an expectation around aligning algo performance to execution goals, which at its most fundamental comes down to long term outperformance against risk transfer pricing.
“Clients better understand the tradeoff between capturing spread on their execution against the market risk they assume, when for instance, making the decision to be passive, neutral, or aggressive in their execution. Pretrade analytics are key to helping clients with these decisions,” Marsden says. He continues: “One of our key algo product differentiators is how we manage and curate our liquidity sources. Clients broadly favour spread capture along with minimal market impact, so the choice of liquidity sources is critical to achieve both those objectives.”
EVOLVING EXPECTATIONS
At ANZ, Marsden explains that the same liquidity sources and algorithms used by ANZ for internal e-commerce hedging also form the building blocks for its client algo offering, meaning that both are objectively aligned. Using TCA by liquidity pool, he explains that clients can visualise and review granular, as well as overall, execution performance. “We believe TCA should always be exclusive and transparent of any brokerage/mark up, so customers understand the raw wholesale market pricing and impact of every individual order,” Marsden says. “Clients expect that we are always sourcing and routing the most efficient, low impact, liquidity on their behalf.”
Clients better understand the trade-off between capturing spread on their execution against the market risk they assume”
Joel Marsden
In addition, the general requirements of nearly every client now includes pre-trade analytical tools with insights into the aggregated electronic market, insights into spreads, insights into the potential cost of execution as well as comparative studies of risk transfer versus algo execution, says Ralf Donner, Head of Marquee Execution Solutions at Goldman Sachs. He adds that most clients tend to also expect some kind of real time order monitor or even a real time order manager, depending on the platform, while post trade they would certainly expect both a post trade report from the bank as well as a menu of third party transaction cost analysis, in addition to some bank aggregated reporting as well.
“Typically, the onboarding process for a new account algo account will also require some kind of evidence of past performance from the algo provider. Any new algo client that we are onboarding now asks us for a study of the most recent set of algo executions, across all currency pairs and what the distribution is, how did they perform? Clients these days will often ask us some really insightful questions,” Donner says. Some of the some of the larger algo providers such as Goldman Sachs now have a dedicated algo team, which Donner notes is not that common. “Particularly in the Asia session, some clients really like the fact that they can speak to an algo expert anytime of the day. That’s part of the personalised service we provide,” he adds.
DEVELOPING INTELLIGENT SOLUTIONS
“However, what I hope we do not end up doing in foreign exchange, and there are some worrying signs from certain hedge fund clients that we are going in this direction, is having to actually create too many bespoke products for clients. A bespoke algo product is likely to be a short term gain and long term pain as it exponentiates testing and thus slows future development.” says Donner. Instead, he says that his team does sometimes need to help clients who are using an API connection via a multi-dealer platform to complete bespoke work on their Fix connection, such as a translation of their settings on their end to the bank’s Fix messaging, for example. “Many of the multi-dealer platforms we deal with are insufficiently on top of this,” Donner adds. “It should be their job to ensure this translation layer is done. It is also in their best interest to ensure that their clients using Fix connections are provided with good service, but some of the smaller multidealer platforms still seem unable or unwilling to do this.”
A further area to explore is the potential for taking a more multi-asset approach to algo development”
Ralf Donner
The increased availability of algo data has also raised client expectations around the outcomes they can expect to achieve using algo execution, says Asif Razaq, Global Head of FX Automated Client Execution at BNP Paribas. Client expectations in the past were more exploratory, but the readiness of analytics and TCA has resulted in those expectations becoming much more focused, he adds. “That is a big change for the market,” Razaq says. “We continue to meet their expectations through the ongoing development of our algo solutions, which have evolved considerably since the launch of our first algorithm almost 10 years ago.” Today, BNP Paribas offers fourth generation algorithms, or interactive algos, which can provide real-time feedback and, using the digital trading system ALiX, can provide commentary to the client mid-execution so the client can modify the algo’s flight path if needed and achieve a much better outcome on the execution. “We have been constantly building more and more tools, which give clients much better control over the execution and a much better overall understanding of how the algos work,” adds Razaq.
Another area of significant development is the demand from clients to build customised solutions. According to Razaq, while the BNP Paribas algo suite is relatively small by design, it should cater for 90% of client execution needs. “However, if a client does have a very bespoke requirement to adapt an existing algo to make it a perfect match to their needs, then we are able to tailor the algo accordingly,” he says. “This framework, Flex, allows us to evolve our algo family without confusing clients with too many different core algos, but instead tailoring the existing algos to specific client needs. We now have around 20 different variants of Chameleon and 16 different variants of Iguana, all designed to do something very specific for those clients’ needs.”
IMPACT OF ANALYTICS DATA
In addition, BNP Paribas has launched fifth generation algorithms, complex strategies designed to perform more than just execution, such as automating the client’s pre-execution workflow. One example of this is Rex, a portfolio hedging or basket algorithm designed for clients who typically do not have just one currency pair to execute, but five or six. “This takes optimising client executions one stage further. Rex can automate the entire workflow, construct a project plan for the client and show them how it can use algorithms in a collective, daisy chain, fashion,” Razaq says. He notes that much of the success of the fifth generation algos has been the
result of the time and effort taken by his team to integrate the algos with the multi-dealer platforms that the clients use. “This makes it much easier for clients to use the algos because they are already integrated to their tool that they use on a day to day,” Razaq adds. “The uptake from clients has been really positive as a result. We’re now brainstorming with clients on other ideas and other complex workflows that they have as an organization that we can look to automate.”
Marsden believes that pre-trade analytics are also becoming increasingly important in educating and guiding clients with their specific execution objectives. As clients assume the market risk on the time taken for the execution, it is imperative that they are equipped with an understanding, on average, how long an algorithm takes for any given time of day or season given prevailing market conditions, he explains. “To that end, the pre-trade analytics which ANZ provide overlay liquidity heatmaps with the recent historical expected trading volume and velocity, which is particularly important for clients that execute more passive or implementation shortfall style algorithms where the time taken to complete is variable,” says Marsden. “Our pre-trade analytics also visually guide clients to be cognisant of known event risk that have the potential to materially move the market, such as news releases, major economic indicators along with daily benchmark fixes, which many institutional clients may otherwise not be aware of.”
According to Donner, there is also much more that can still be done to increase transparency around how the algos with different liquidity sources. “Algo analytics feels like it now pops up everywhere,” he explains. “Yes, there is a huge amount of analytics out there, but it’s daunting. It is like stepping into an aircraft cockpit and there are dials, wheels, bells and whistles all over the place. It’s very difficult to make sense of the entire the entire thing. Clients are looking for something that is a little bit more streamlined, something that shows you everything at a at a glance.”
NEW INNOVATIONS AND DEVELOPMENTS
Achieving this the main focus which will help clients in terms of the usability of the algo analytics, says Donner. He adds that analytics tools also tend to be very disparate. “There will be one pre-trade tool, another separate intra-trade tool and then yet another for post-trade,” he says. “Harmonizing those tools is an aim worth pursuing.” A further area to explore is the potential for taking a more multi-asset approach to algo development, Donner says. From a client perspective, he explains that FX does not sit on its own but, particularly from a client perspective, it sits together with other tradables.
“Client expectations in the past were more exploratory, but the readiness of analytics and TCA has resulted in those expectations becoming much more focused”
Asif Razaq
Another recent release is a new Franchise Matching algo, which was developed as a new way of leveraging internalisation to fill the algo order. Donner explains: “This is all about liquidity. Liquidity used to be a bit of a challenge with algos. There were only a handful of venues and there was a great deal that had to be done externally.” He adds that algos in general have improved over the years and have become much smarter about how they source internal liquidity. “Our new Franchise Matching algo is yet another example of being up being smart about sourcing internal liquidity,” Donner says. “It encourages the opposing interest, as distinct from previous versions of internalisation which sought existing opposing interest. Basically, it’s a new form of internalisation.”
Razaq adds that at BNP Paribas, they are already ahead of the curve and are recognised as innovators in the algo development. “We’re still enhancing our strategies on a day to day basis and we’re looking at more fifth generation algo opportunities to build upon,” he says. “But we now feel we’ve conquered FX in terms of our capability. The bigger opportunity for us is to now look at how we can apply the FX algo technologies to different asset classes.” This involves building the FX family of algos, Chameleon, Viper and Iguana, but making them available to trade different asset classes, such as commodities, precious metals, and Futures with underlying FX, bonds and equities, adds Razaq. “We are planning to launch these new algos later this year. Our focus now is how can we apply the same principles and the same technology in different asset classes, giving the end client a unified view of algos. Because if they know what Chameleon does in FX, then they’ll know what Chameleon will do when it comes to trading gold or when it comes to trading futures,” he says.
According to Razaq, that common understanding of how the algo works, using the same interface to keep the look and feel of the FX algos, as well as having Alex to provide algo users with commentary on their futures execution and FX execution, will bring valuable unity within the algo offering. “The cost pressures we are seeing now in the world, which our clients are feeling as well, mean that everyone’s being asked to do a lot more with a lot less. Clients are no longer just an FX execution desk, but they’re increasingly becoming multiasset execution desk. The more we can do to help them to align our products to those different asset classes, the easier we can make life for our clients,” Razaq concludes.
