As has been well documented, the Covid crisis served to highlight the stability and effectiveness of algos to traders and opened up demand from a number of non-traditional algo users. Vivek Sarohia, Global Head of Alternative Execution Services at HSBC, says that the success of these products has also meant the need for a deeper understanding, and ultimately an increase in sophistication, from clients around the full life cycle of an algo. “Clients are no longer just placing their algos with their provider and awaiting their outcome, but rather they’re wanting greater transparency on their execution,” he adds. Instead, clients are now pressing their providers to demonstrate a much wider range of factors according to Sarohia, including helping clients to better understand what algo strategies are available, how they will perform based on historical performance and how to know which algo strategies would be appropriate for their execution needs. He notes that clients are also increasingly interested in learning how the algos inherently work, how the strategies will adapt and perform in real-time as market conditions change as well as asking for the ability to amend orders in-flight should those expectations not be met.
Ralf Donner, Head of FICC Execution Solutions at Goldman Sachs, says there has also been a notable overall growth in client expectations when it comes to intra-trade services. He explains that in the early days, before the emergence of third-party TCA providers, clients expected the banks to be able to produce something in house to provide them with some form of post trade debrief. Then once post-trade analytics were more firmly in place, Donner says that clients began looking to algo providers for some form of pre-trade analytical toolkit in order to make better selections around algo selection. He adds: “The expectation from clients now is for the ability to make meaningful decisions during the lifetime of the execution, including how to make informed choices about amending their algo orders.”
Technology norms
The drivers behind the evolution in client expectations is down to the increasing availability of such tools, according to Donner. “Think about driving – a long time ago there were no navigation systems and we still all managed to get from A to B. Now, even though we may already know the route we want to take, we will still turn on our navigation systems because they are smarter when it comes to finding a way through traffic,” he says. “The same is true when it comes to algo tools. Once clients have experienced the benefits of being provided with really useful information about market conditions, then they can’t do without it.” In a similar way, the benefits to clients of being able to customise different parameters for their FX algo orders, such as speed, liquidity etc, have led this to become an expected level of functionality. “It may be common knowledge already that a certain aspect of the FX markets is difficult to manage, such as a typically low liquidity period during the course of the day. But how that problem manifests when trading algos and how the algos are able to deal with that problem will vary,” says Donner.
He explains that FX algos can typically deal with this type of problem in two ways. Firstly, the algo logic itself might have already been built to automatically adjust to market conditions, or secondly, for the client to be able to intervene in order to optimise the algo execution in the case of different or unusual market conditions. “Both are valid, so we offer both,” Donner says. “Algos now have what is almost self-learning logic built in, which tries to accommodate for whatever new conditions may arise, but we also try to give clients the ability to change things themselves. That may be as simple as changing a limit price to deal with market conditions, or it could be as complex as changing the liquidity makeup that is permitted for the algos, to everything in between.”
Flexibility and control
In addition, FX algo providers are increasingly developing new ways to provide clients with greater transparency of the underlying logic powering their algo trading executions. Sarohia says that HSBC has also recently responded to this demand by launching improved pre-trade and real time tools, with the aim of improving the algo user experience. He adds that the new interactive FX analytics suite will empower clients to make algo execution parameter decisions pre-trade in addition to proactively being able to monitor their algo performance in flight. Focusing on the pre-trade component, Sarohia adds that the analytics suite can now provide clients with access to both externally aggregated and HSBC proprietary market data on a number of metrics including volatility, volume, market depth, spread and upcoming news events. “In combination with HSBCs own proprietary algo simulation models, which have been made open access, our clients can now simulate outcomes from all of the algos before they execute,” he adds. “This gives them the ability to vary execution parameters, from algo strategy, to leniency, to time of day and identify with confidence the most appropriate algo strategy for their trading needs based on historical data.”
Clients are no longer just placing their algos with their provider and awaiting their outcome
Vivek Sarohia
However, Sarohia adds that historical predictions can never be perfect indicators for future performance and warns that market conditions can change very quickly with little warning. He explains that these concerns can be particularly acute with algos, where the perception of the ‘black box’ deciding how the algo should execute in these scenarios can become less appealing to certain clients. “In order to address that particular issue, we have developed ‘in-flight’ assessment of the algo order through our ‘real-time’ analytics component,” Sarohia says. “Clients are given access to how their algo is performing, both on an absolute level in terms of price and fill, but also against various industry standard execution benchmarks, all in real-time. In addition, clients are provided with the ability to drill down into each underlying child fill, thus being able to immediately identify in which venues the algo is executing, its market footprint and whether the algo is meeting its expected outcome.” In conjunction, Sarohia notes that the analytics suite is also combined with an algo order tile, allowing the client to adjust their order exactly how they see fit across every order parameter, with the amendment being able to be actioned immediately. He explains that this includes having the ability, should clients believe it is necessary, to close out the order immediately via risk-transfer price using HSBC’s own liquidity stream. “This flexibility provides assurance to the client that they are always in control of their execution and provides them with more than one method to execute their needs,” he adds.
Information at their fingertips
Overall the demand for the ability to customise FX algo executions has been fairly consistent, according to Donner. However, he adds that providers are increasingly looking at the general trend towards a reduction in the share of the liquidity pie that primary markets now hold in light of the migration to other sources of liquidity. “The demand for selecting liquidity comes to some extent from the fact that we’re no longer in this old paradigm of having internalisation plus primary and then some other liquidity sources, but actually now, it may make sense to switch off the primary source or to use other maybe softer mark-out liquidity instead,” he adds. “Clients increasingly want to have more transparency around how their algo orders interact with different liquidity sources.” Donner notes that this can take a number of different forms, including the standardised client disclosure templates offered by the GFXC following the revisions to the FX Global Code or those offered by the Investment Association. “Clients ask a lot of questions about how exactly internalisation operates, including what our definition of internalisation it or how do we prioritise in terms of franchise liquidity versus other forms of liquidity? We try to get this into writing to make sure it is as clear as possible, which the Global Code standardised algo cover sheets is expected to help with,” he says.
Goldman Sachs has also developed its real-time order monitor, which has been available to clients for some time now, adds Donner. It shows clients a variety of market charts such as bid/offer spreads or fills colour-coded by venue. “Colour coding and other visualisation tools are useful in these diagrams because clients can see everything at a glance,” Donner says. “For example, if the chart of fills is mostly purple, they know they are mainly getting internal fills. So it allows clients to make much faster assessments and decisions.” In addition, clients can plot parameters related to the algo, including ‘no worse than’ or knock-in and knock-out levels. Clients can then use that data to decide to pause the algo, or to knock-in an algo or to fully fill an algo, depending on what they plan to do, says Donner.
Finding improvements
“Another tool which clients find of interest but which is hard to achieve with FX is to be able to see how their algo execution compared to some rough estimate of the market as a whole,” Donner adds. “So what’s the market typically doing around this time in terms of executed volume, on this particular day. We try to present all of that information as well.” In addition, he says that clients also like to have live updates of algo performance vs benchmarks as well as being able to see at a glance the nature of the liquidity or to get a sense of whether there will be any undesirable markouts from a given venue. “This can all help the client to decide if they want to switch liquidity groups, or if they want to pause the algo or turn it off if it looks like it’s creating market impact. So the real-time order monitor is the first point of call when it comes to launching an algo with us,” Donner adds.
The expectation from clients now is for the ability to make meaningful decisions during the lifetime of the execution
Ralf Donner
Looking ahead at how to further improve the client experience, Donner says that the demand is mainly for the bank to continue reviewing and improving the toolkit to ensure it is the best available in the market. He explains that a good example of this is the duration estimator tools which most providers offer for FX algos. “However, we regularly hear that the models out there are quite poor at estimating algo duration in reality and so we’re currently researching ways in which we can improve that,” Donner says. “We are working on a brand new duration model that uses a Markov chain model to estimate transition probabilities effectively based upon historical child orders of algos, instead of the current approach which is typically based on historical parent order executions.” He adds that this new approach to modelling algo duration is likely to prove more accurate that the current models used and expects it will be made available to clients late this year. In addition, Donner adds that much of the recent focus has been on developing pre-trade analytics for FX algos, but notes that FX does not exist in a bubble and typically leads or lags behind other markets, such as commodities or rates. “There’s now growing interest in having a dashboard that is cross asset in nature and which looks at looks at market microstructure as a whole, not just for FX but for related asset classes as well,” he says.
Next generation developments
In contrast, Sarohia believes that the increased interest in tools powered by machine learning or AI will be key to further improving the client algo trading experience. He explains that the bank has been at the forefront of AI within the FX space through the launch of Sympricot, a chatbot which uses Natural Language Processing (NLP) technology. Sympricot has been providing clients with instant pricing and analytics for FX options since the start of the year. “In our view, NLP is one of the most natural ways of interaction and having the chatbot interface allows HSBC to provide lots of added value to clients,” Sarohia says. “We have continued to expand Sympricot’s use across other areas of FX and, in conjunction with the launch of our FX analytics suite last month, now allow clients to utilise Sympricot for algo execution risk management, providing access to in-depth information on liquidity conditions and expected pre-trade algo performance assessment via the bot.”
He adds that it most notably gives clients an immediate way to access information about the FX markets, which previously would have been difficult to obtain quickly, and allows a multi-product user of FX to connect seamlessly to HSBC’s infrastructure and knowledge base within a unified framework. “We see Sympricot leading to more interaction, better dissemination of information and ultimately more transparency,” adds Sarohia. “However, we believe that it is only the first step. The continuous development in electronic trading and HSBC’s expertise in machine learning for systematic trading will naturally lead to the manufacture of high quality analytic services to be distributed to our clients. For algos in particular, we expect the future of the service leading towards creating fully customisable algo strategies and analytics, ultimately providing even more personalised and bespoke outcomes for clients.”
