When algos first came to market, they were supporting trading in the most liquid markets, the G10 currency pairs, as it was felt that these were the easiest markets to handle when it comes to building algos, says Asif Razaq, Global Head of FX Automated Client Execution at BNP Paribas. The next step in expanding the product coverage of algo strategies was to start building dedicated algos to trade in EM currencies, which according to Razaq was necessary because those markets function in a very different way to the liquid G10 and involved a completely different data set. “The same logic applies to our current development of purpose built strategies to trade the NDF markets,” he adds. “These algorithms are very different because again, the underlying markets are very different, both in terms of how we source liquidity, but also because they trade against a one month tenor. When providing algos in this space to support crosses, you also have to trade the cross currency pair with the same value date. So we decided to start from the ground up with the build of our dedicated NDF algos and the performance of these strategies as a result are very strong.”
In addition to expanding the product set to include new markets and currencies, Razaq says the bank is also currently looking at introducing additional workflow optimization tools. “Our ethos was always to keep the algo family small. So for the last eight years, we’ve had the same three core execution algos which we’ve been tweaking to continually improve their performance,” he adds. “We have now added two more algos to the suite. These algos were more specifically designed to tackle not execution – in the same way these initial strategies were – but to tackle particular workflows.” The first of these new algos to be released was the DMA gamma hedging algorithm, which Razaq explains was designed to help clients manage the delta hedging of an options trade. The gamma algo utilises the same technologies of the core execution algos: Chameleon; Viper and Iguana, to manage sub-algos which execute the delta hedging of the client’s gamma positions.
The second new algo which is in the process of being built is Rex, which Razaq explains will be a new strategy specifically developed for portfolio hedging. “In a similar way, we’re again utilising a master algorithm which is then controlling the existing algos to decide how you execute a portfolio of currencies or portfolio of exposure. This is a key area of focus for us at the moment.”
Deutsche Bank has also been working on further enhancements to its existing FX algo offering, including the recent addition of a new product called Principal Resting Order (PRO) Liquidity, which expands the number of liquidity pools that the algos can access, says Vittorio Nuti, Global Head FX Algos at Deutsche Bank. “We have started to intelligently work orders via the franchise through the use of PRO orders and that now forms quite a substantial amount of our overall algo execution,” Nuti says. He explains that this creates a number of benefits to clients, such as being able to cross interest between the algo clients and their orders.
It is more pertinent than ever that providers and end users review their chosen execution algorithms
Improving execution quality
“We’re already seeing north of 20% of the flow in our G5 currencies going through PRO. This addition has created a substantial decrease in execution times and costs for our algo clients and we plan to continue improving these capabilities, including our mid prediction,” says Nuti. He adds that there has also been a recent emergence of new venues in the peer to peer or mid book arena, but believes many models still have issues. He explains that they tend to either offer some form of blended mid, which he warns can’t really be relied on, and/or the venues add their fees to the fill reported as trades, making it too expensive. “We’ve noticed that the cost of using some of these new liquidity venues were even higher than the actual bid offer that we would have to cross in some of the pairs, so it made little to no sense to actually use them,” Nuti adds. “It should be about accessing liquidity smartly by reducing adverse selection, ensuring that you don’t cross a large bid offer while executing, and that can only really be achieved by superior mid prediction.”
“We want the end user to be plugged in and involved in the execution – if they want to delegate everything to the algorithm, the machine and the algos will be fully capable of executing the entire transaction..” – Asif Razaq
In addition, John Burke, Head of eFX for Northern Trust Global Foreign Exchange, says that currency markets during the pandemic showed a lack of deep liquidity, alongside the heightened volatility, coupled with an underlying tone of increased market fragmentation and warns that although some return to normality has occurred, it is more pertinent than ever that providers and end users review their chosen execution algorithms. He adds: “We are seeing increased adoption of algo execution as part of a continued trend toward FX outsourcing in response to cost pressures and increased comfort, with traders providing oversight to algo executions rather than taking up valuable resource availability.” While time based or fixing algorithms and orders will always remain a large part of the FX market, particularly for securities service clients who require fill certainty, Burke believes there is a notable shift in the types of algorithms being employed, with clients increasingly interested in algorithms designed to actively source good quality liquidity at assessed price levels. “These hybrid algos offer flexibility, while still ensuring fundamentally required characteristics are met, whether it be fill certainty, limit levels, etc,” says Burke. “They may also have an underlying grounding in time, visibility, estimated market volume or other external factors and can actively seek to provide an incremental increase in value through targeting optimal executions across liquidity providers or venues.”
The second- and third-generation algorithms which are now emerging seek to incorporate some level of intelligent decision making into the process, or are at least dynamically responsive to market adjustments during executions, Burke explains. The continued impact of MiFID II and the need to achieve demonstrable quality execution has also encouraged market participants to question not just their method of achieving a good face-value rate, but also to look further into the underlying risks and costs associated with trading, he adds. “This has been a key driver of FX outsourcing, not just to control operational costs and efficiencies but to achieve scale and access to liquidity,” says Burke. “The key to effective oversight is the relationship between all parties involved in an algorithmic execution – it should be open and transparent and ultimately lead to providing comfort to the asset owner or manager. Pre-trade analytics and liquidity management should therefore be considered extremely important tools when assessing algo execution.”
Putting the client in control
In a quantitative sense, Burke believes that TCA should be encouraged so that clients can understand not only the metrics around measuring market impact and slippage costs, but where, how, and why any particular sub-order has executed as seen. For example, service providers should be able to provide comfort to clients around the additional hold time policies within last look statements of execution panel participants, he adds. “We also believe service providers need to justify to their clients that there is no potential conflict if they plan to include their own liquidity in the panel. These discussions are key to providing a holistic execution service and can be crucial to building a tailored TCA program to suit the operating environment for individual clients,” says Burke. “For Northern Trust clients, TCA is becoming more of a value-add, with data harnessed to achieve better outcomes. Advances in reporting, based on iterative discussions, continue to improve transparency and the user experience for TCA clients.”
“Sourcing liquidity and continuing to invest in top class systems and infrastructure is going to be key for helping clients to go through the next phase” – Vittorio Nuti
In addition, clients increasingly want to have more control over which venues they want to trade on and in some cases they may prefer to only trade to one venue exclusively, Razaq says. In response, BNP Paribas’ algo platform has the facility for clients to either delegate the execution choice of venues to the algo, or they can opt to cherry pick which venues they want to trade on. “In an illiquid landscape we’re also seeing a slight shift in how we ourselves approach the market,” Razaq adds. “Mid matching venues have emerged as an increasingly popular option with clients and so again, we’ve re-engineered our algo strategies to focus more effort in finding these mid match opportunities, rather than going to the lit market.” He adds that the rise of peer to peer venues also offers another way for buyside clients to show the order less to the market and to create less noise, making the pool as safe as possible in terms of reducing information leakage.
A further key focus at BNP Paribas is the development of interactive, or fourth generation, algorithms. Razaq explains that this moves away from the concept of algos as just a black box to demonstrating that algos are actually an interface with which clients can interact throughout the execution. This includes the addition of applications such as Execution Manager, which Razaq says enables clients to see for the first time the live performance of an algorithm in flight, while additional tools through Insight Live put clients in the driving seat to give them full control over how they interface with the market. “But more importantly, we’re now giving them the visualization of all the underlying data so they can actually make an informed decision, as opposed to a blind decision which in the past was their only option,” Razaq adds.
Quality over quantity
“The key to effective oversight is the relationship between all parties involved in an algorithmic execution – it should be open and transparent…” – John Burke
At Deutsche Bank, Nuti says the focus is on building tools for the Market Colour app, which will enable clients to view and amend orders via the app, which is typically accessed from Autobahn but can soon be accessed from Bloomberg regardless of the OMS used to input the order. “Clients can then open the ticket and do all the amendments that they wish, there is a wealth of parameters which we will be releasing before the end of the year,” says Nuti. “In addition, the bank has introduced straight through processing in terms of passwords etc which enables clients to log in seamlessly from Bloomberg into our system and view their TCA straight off in real time, which can be very useful and powerful.”
Nuti says it is more important than ever that providers continue to invest in execution, particularly as market liquidity is tailing off slightly and margins continue to be razor thin. “Sourcing liquidity and continuing to invest in top class systems and infrastructure is going to be key for helping clients to go through the next phase,” he adds. The last 12 months in particular has also seen a massive move to the electronification of NDF pricing, adds Nuti. “Liquidity is now deep enough to execute decent sizes in relatively short periods of time, sometimes even faster than some of the more illiquid emerging market deliverable pairs,” he says. “Our newly introduced Brazil NDF is fairly liquid and we now have all the EM, all the LatAm, we have Brazil BMF1 and BMF2 expiry dates and we also have IMM1 dates for the more liquid Asian NDFs.”
In terms of future algo development, Nuti says that his slogan is fewer but better, and aims for a streamlined algo offering which offers improved execution and simplifies the selection process for the client. “As a result, we’re focusing on merging our existing algos and we’ll be reducing further the number of algos that we provide to instead concentrate more on delivery of the particular execution style that the client wants to use,” he adds. “We are in the process of also releasing a whole new technology underlying our algos, which will have the algo engine sitting a lot closer to co-location. We’re also introducing a better predictive algo, which will make use of this new technology and will include all the best features of our existing algos.”
Automation and algo growth
The second- or third-generation algorithms now emerging also seek to incorporate some level of artificial intelligence, or at least are dynamically responsive to market adjustments during executions, according to Burke. “We are bringing new algos to market that are based on intelligent decision logic. These algorithms are designed for use in an outsourced FX approach, without sacrificing transparency or execution quality,” he adds. In terms of what more leading providers should be doing to support clients, Burke says there should be a strong focus on clear and proactive communication between client and service provider. “This is true at every step in the process, from the choice of algo to achieve a specific result to transparency around transaction cost analysis. For example, good communication will provide clarity around the additional hold time/last look policies of execution panel participants or help to explain the inclusion of an algo service providers’ own liquidity into the panel,” he says. “These discussions are key to providing a holistic execution service and building tailored programs suited to the client’s operating environment.”
Algorithms are now emerging that seek to incorporate some level of artificial intelligence
Looking ahead, Razaq believes that the further development of AI and machine learning tools is a key growth area and a vital area to invest in if providers want to build smarter algos. “We’ve always liked the hybrid approach, so using AI to provide the algo with the intelligence to navigate those markets, but while retaining that human touch element,” he adds. “We want the end user to be plugged in and involved in the execution – if they want to delegate everything to the algorithm, the machine and the algos will be fully capable of executing the entire transaction – but we believe that interaction provides the client with more colour on what is happening in their execution.”
Razaq believes there is more scope for future developments based on AI and machine learning, as well as more work to be done in supporting algo services for trading outside of FX spot. He adds that the next new emergence will be in the development of algos for the swaps market. “As the swap market becomes more electronified and we see more venues popping up offering swap liquidity, we can now expand the algo universe to also trade FX swap products in line with the FX spot algos,” Razaq says. “That’s a very important development that we’ve seen that should improve the whole execution experience. Most of our clients have exposure on the outright position, but at the moment they have no means to algorithmically execute that swap component of the outright.” He adds that a second area that the bank is looking at is how to develop tighter integration for the end clients with their OMS platform. “At the moment, end clients are having to manually launch these algorithms through an OMS platform of their choice. We now want to automate that submission by directly bridging to the OMS platform that the client uses. The more they can automate on their side by automatically launching these algorithms and automatically processing the TCA and analysing and scoring algorithms, the more adoption we’re going see in this space,” Razaq concludes.
