Buyside participants have been increasingly gravitating towards FX algos since their experience of market volatility at the start of the pandemic, which saw some liquidity providers and market makers stepping away and FX liquidity became a lot thinner, says Tan Phull, Head of Global FX Execution Trading Services at Bank of America. Since then, the FX algo market has been evolving and consolidating, Phull explains, with most of the strategies available today now falling under a few key categories. “Most algos today generally fit within one of three buckets: schedule based algos, such as TWAPs and VWAPs; participation algos, such as arrival based or volume based algos; and the final group, which are the more aggressive, liquidity seeking algo strategies,” adds Phull. “What we’re trying to do now is to increase the refinement around how the algos work within those buckets to really simplify and streamline the algo experience for the buyside trader, while also providing access to a much richer set of features and a deeper amount of liquidity when using them.”
“Liquidity access, data and customisation underpin the effectiveness of an algo suite.”
Tan Phull
Clients are also becoming increasingly sophisticated in their use of analytics and data in order to measure the effectiveness of an algorithm, according to Phull. In order for his team to also have a better understanding of the performance of the various algo strategies they offer, they recently partnered with third party TCA providers and uploaded all available algo data for last year as well as for year to date. The aim, he explains, is to work with these providers to create and independent measure of the bank’s algo performance according to their metrics. “We wanted to measure ourselves in the same way that our customers would measure us, then to use those findings to significantly improve our FX algo offering and invest in a mechanism will further help our clients,” Phull adds. “One key area where we were shown to be particularly strong was in our liquidity seeking algo, Whisper. This is what innovative, next generation algos focus on, providing better access to unique liquidity, be that through internalisation or external liquidity pools, and using the available analytics data to help customise the algos according to the way they want to trade.”
Meanwhile Nikhil Joshi, eFX Sales at NatWest Markets, adds that due to their simplicity, TWAP and VWAPs have probably been the most popular algos for those who are looking to test performance and they were a good starting point to replicate working an order manually over a period of time. “They are, however, also fairly limited in how they executed – with a TWAP you look to slice by time, notional, number of slices etc but the sophistication ends there,” he adds. “VWAPs, where you are looking to participate in the market against a historic volume measure, such as external interbank liquidity, is also fairly simple by nature.” On the other hand, Joshi notes that smart order routing and intelligent placement algos have become more popular as markets have become further fragmented and the focus has moved to avoiding any market impact, if at all possible.
Additional capabilities
“It’s not just how algos slice and place orders, what is equally or even more important are the pools they interact with,” he adds. According to Joshi, order placement and routing logic become almost irrelevant if you are placing into a pool that caused immediate leakage and impact, therefore having bespoke curated pockets of liquidity is ever important. One aspect of these order types is the fixed timeframe which is specified up-front, he explains. While this provides the benefit of timing certainty, the likelihood is that the algo user will end up crossing spreads for execution. “Using algos that only place resting orders, like the NWM Peg Clipper, which fill as quickly or slowly as the market activity dictates, will invariably yield better results over the longer term,” Joshi says.
“It’s not just how algos slice and place orders, what is equally or even more important are the pools they interact with.”
Nikhil Joshi
Ralf Donner, Head of FICC Execution Solutions at Goldman Sachs, says that rather than traditional TWAPs/VWAPs, the most popular algo strategies that his team offers are actually those which do not have an end time to them and do not predetermine the strategy window. Instead, next generation algos go with liquidity and the client is able to choose the parameters to control how opportunistic it will be, how much market risk it will take, whether it is a very passive or very aggressive strategy or somewhere in between, he explains. “That’s what we normally see as being the most common and popular strategy,” Donner says. “Nothing much has changed there, although one thing that we have seen over the last couple of years is a growth in the usage of purely passive strategies.”
According to Donner, it used to be the case that if you were only posting and you were not ever aggressing, then you wouldn’t be able to fill in a reasonable time and it would be a struggle to complete execution. “But a number of things have changed since then,” he says. “One is there’s less reliance on primary markets, so there are more idiosyncratic pools of liquidity that can be called upon such as internal matching engines, but also mid pools and other bilateral sources of liquidity. And then the second important change has been the increase in useful algo analytics, particularly pre-trade. Now, before somebody inputs a passive strategy, they’re more likely to go and check in on our analytics platform to see whether it makes sense at that time of day and in that currency pair to use a purely passive strategy. That in turn increases the confidence in using that strategy and so we actually see more and more very passive strategies being executed as a result.”
Driving greater customisation
In comparison, Donner adds that just two years ago, the bank’s main flagship strategy was its dynamic hybrid algo, which he says was where nearly every client wanted to be executing. “Now we see large clients beyond just the corporate client base, such as hedge funds and real money clients, using the purely passive strategy more as well. That is a significant change.”
At HSBC, the continuing focus is on enhancing liquidity pool curation, says Vivek Sarohia, Global Head of Alternative Execution Services at HSBC. He explains that this includes the creation of bespoke pools of liquidity across HSBC’s ECN network to reduce post trade market impact (and ultimately tighter spreads and higher fill rates) in addition to smart allocation to improve the decision making of the algos to drive where fills are placed, which improves fill rates and again reduces market impact. “Our most popular algos have been our TWAP, Liquidity Seeking and Implementation Shortfall Algos and represent nearly 90% of algo orders placed with HSBC in 2022,” Sarohia adds. “We continue to observe that clients are requiring more and varied different execution metrics and pushing for enhanced TCA requirements and/or third-party involvement to analyse their algo performance, alongside pre-trade analytics for informed risk management. Especially with the increase in volatility seen in the markets over the past few months, the importance of having dynamic control of your desired execution outcome has taken greater importance.”
Within an algo suite, the strategies will come with their pre-defined set of parameters and this default setting may not suit every client exactly, notes Phull. “We need to do work with the client as an execution consultant to help them tune the algos capability or settings according to the way that they trade, according to the way that their funds are managed, according to the way that their investment process works. Liquidity access, data and customisation underpin the effectiveness of an algo suite,” he adds. For example, to further enhance liquidity access, Phull says that the bank has invested in making sure that its internal liquidity franchise is available to algo users in the most effective way. In addition, the provider has also been enabling access to innovative types of liquidity, such as from ECNs or peer-to peer networks.
The buyside is becoming increasingly focused on execution costs
Fine tuning to improve execution
In addition, Phull says that the buyside is becoming increasingly focused on execution costs. This is being driven by both best execution requirements and also because the buyside is increasingly becoming more sophisticated in their use of data, he explains. “This gives us the opportunity to work with our customers and have those conversations based on the analytics, but then the data also enables us to quantify the improvement that they can gain from that customisation. It’s an interesting and exciting time because as the clients become more sophisticated, this creates more opportunities for us to work increasingly closely with them to hone their execution performance,” says Phull.
As a result, he explains that the key area of focus for the further development of the algo suite is to continue simplifying the algos while also further enhancing the amount of liquidity that the algos can access. In addition, due to the size of Bank of America’s franchise, Phull says an ongoing major project is to ensure that all available liquidity from the many different business areas is being piped into one central liquidity source where is can be accessed by the algos. In addition, the bank is in the processes of rebuilding all of its existing algos using a new common infrastructure for both equities and FX, which Phull says will significantly boost the capabilities of the FX algo suite as a result. “The objective of this exercise is to take our existing algos and really revamp, which will in turn help us to make them even more customisable, to further improve the way that they access liquidity and overall, to significantly benefit our clients in achieving their execution outcomes.”
Customisation
“Now we see large clients beyond just the corporate client base, such as hedge funds and real money clients, using the purely passive strategy more as well.”
Ralf Donner
When it come to the demand for FX algo customisation, Joshi says that while requirements can vary considerably, buyside firms are often motivated by the need to alter the speed of execution on a case by case basis or to operate in a specific set of liquidity pools. “This latter point is particularly interesting as the motivations do vary widely and can be both policy driven and performance driven,” he adds. “Some examples from a policy perspective would be wanting to use no-last look liquidity only, or perhaps have an algo which will only interact with market participants who have a certain ESG rating. Such choices inevitably affect the rate of execution and one of the potential pitfalls of any customisation along these lines is the inevitable conflict between wanting quick execution in a more limited pool.”
In addition, speed and impact are highly correlated, according to Joshi, who adds that it is appealing to many users, and algo providers, to offer an ever increasing range of settings on an algo. However, he warns that the challenge then is how to judge performance in a statistically meaningful way, when every new setting dilutes the number of executions that have that particular combination. “‘Peer group analysis’ being offered by various third party TCA providers can be a big help here and greatly improve the reliability of such analysis. We would highly recommend users take advantage of this data, but despite this there are clearly diminishing returns to adding more and more complexity,” Joshi adds.
He explains that the market appears to have settled on having a choice of three of four ‘speeds’ that the user can choose from, as well as offering a limited selection of liquidity pools. “In our experience going from the slowest to fastest can easily generate a factor of 10 increase in speed, which is more than enough for the vast majority of use cases,” he adds. “That there will always be more market impact the faster you go cannot be avoided and we do tend to steer users to go slowly, unless they have strong reasons for speed.” With that in mind, he says that the focus for NatWest Markets going forward will be on curating better liquidity and identifying counterparties that cause market impact. “This is a moving target and is and one that requires further data mining to reach better outcomes, and significant investment in machine-learning capabilities, however is the best route to improving the outcomes from an implementation-shortfall perspective,” adds Joshi. “We have seen recently during high times of volatility and thin liquidity, that you need strong liquidity providers and pools for your strategy to interact with, and these need to always be ones where information leakage is under control in all circumstances.”
Ongoing growth and future development
“We continue to observe that clients are requiring more and varied different execution metrics and pushing for enhanced TCA requirements and/or third-party involvement to analyse their algo performance.”
Vivek Sarohi
For Donner, the most important area in terms of algo development is the provision of useful algo analytics, particularly pre-trade. “This remains a big focus for us. We are aiming to build an analytics offering that is not only robust but also cross-asset in nature. So the aim is to have something that would work not only in FX, but also in other asset classes and which actually wouldn’t just be about algos. “Of course, the work on algo development is also ongoing. That remains an arms race and we are very, very busy there. But we expect to see more in terms of the unique interactions between analytics and algos, this will be an important difference for the algo market.”
Looking to the future of the FX algo suite at HSBC, Sarohia shares the news that the bank will providing clients with the ability to change strategies ‘in-flight’, a new offering which it will be launching in the second half of this year. He adds that this will allow clients the flexibility to switch between HSBC’s aggressive and passive algo strategies during the life of the order, without having to cancel and rebook, thus allowing clients to speed up or slow down execution as market dynamics change. Also, in conjunction with the bank’s algo analytics suite, Sarohia says they are looking to expose more of their proprietary risk management tools, which can allow clients to have greater confidence on the likely range of outcomes of their execution.
Finally, internalisation is also often quoted as a key differentiator from the sell side regarding reducing market footprint, but how internalisation actually takes place varies across institutions, says Sarohia. “At HSBC, we are putting a lot of analysis and modelling into how clients can internalise their flow across HSBC’s various client types,” he concludes. “Careful consideration is clearly needed regarding information leakage, but the opportunity for clients to dynamically choose how their algo orders offer to, as well as take, opposing flows of liquidity from different client sectors, especially from a firm such as HSBC with its global footprint and diverse client base, has extreme appeal and could be the next step in terms of driving tighter spreads and higher fill rates. Ultimately, transparency is key and analytics, combined with flexible execution tools, helps achieve our clients’ goals.”
