Nicola Tavendale

What lies in the road ahead for FX algo trading in 2024?

December 2023 in Industry Views

One constant in any market is change. Looking back over the past year’s headlines in FXAlgoNews and it is easy to see that the evolution of the algo industry has continued apace. The market is more mature and clients are increasingly adopting algo use for a wider range of order types. Meanwhile across the wider FX industry, the impact of changes to volatility, liquidity access and the arrival of new venues and technologies is also being felt. Some of world’s largest banks are spearheading a great deal of the current innovation that is taking place in e-FX so what developments do they expect to see with algorithmic FX trading in 2024? Nicola Tavendale investigates.

Following three years of gathering insights from an annual Global Client Fixed Income Market Structure Survey, Barclays conducted a similar exercise with its FX clients for the first time this year. The results provide a telling insight into the key areas of focus for clients, particularly so for the FX algo business. One of the most significant findings were the number of clients who were planning to look more closely at FX market liquidity and how to access it, says Ajay Kataria, Head of Electronic FX Distribution, Americas at Barclays. “For our clients, liquidity access was shown to be their biggest area of concern,” Kataria adds. “We believe this is also a big part of the story for 2024. Putting this in the context of FX algos, this means the focus for our clients will be very much around finding the best liquidity and navigating this changing market structure and landscape.” 

When clients were asked how they review a dealer’s algo functionality, the most important area was found to be internalisation. Kataria explains that internalisation will also continue to be a key area of focus for the business to further leverage the benefits that come from having such a large franchise. Ensuring that the FX algo offering is constantly improving and adapting to the changing FX landscape it also essential, says Preston Mesick, Global Head of FX Algos at Barclays.  “This will be a big area focus for us next year,” he adds. “FX is very much a follow-the-leader market. With new venues and new orders types coming to market, this will really change the nature of FX liquidity.” Being able to help clients navigate the changing landscape and market structure will be very important looking to 2024, notes Mesick. “For example, as the market and NDF liquidity expands, we will continue to assess and update the number of liquidity sources that our NDF algo offering is able to access,” he says.

Another big theme for the coming year will be improving the quality of FX algo execution assessments, according to Vittorio Nuti, Global Head of Listed Derivatives & FX Algo Trading at Deutsche Bank. “The industry as a whole is maturing and the algo clients are becoming increasingly sophisticated. With that comes demand for a better quality assessment of the algos that they are employing, whether that is through third-party TCA or their own in house analytics,” Nuti says. Assessing the provider of the algo should also be an important consideration, he adds. “The quality of execution will naturally vary over the course of weeks, months or even years, which is why it is important to assess the quality of algo execution at a global level,” says Nuti. “For algo providers, that assessment will be key to getting more business.” 

Increasing competition

A further trend which is increasingly coming to the fore is demand for client customisation of algos, although as Nuti explains, it can prove to be very difficult to customise algos to clients requirements. He says: “Most of the time, those requirements are quite fuzzy in nature. This is why the industry as a whole is based on results. It is about achieving the right outcomes for the clients, but ultimately each client may have slightly different targets that they’re aiming for in their algo execution.” At Deutsche Bank, the focus is in on having these detailed conversations with clients to understand how to help them effectively. “One of the main strengths of our algo business is that the team is perfectly aligned with wanting to achieve the best outcomes for our clients,” says Nuti.  

Meanwhile at Goldman Sachs, the recent development of a new way to bring internalisation to algo execution is also notably gaining traction. Called Franchise Matching, it works in partnership with the ebook to create an offsetting position for the FX algo, even when there is not one immediately available. This has been particularly successful with more sophisticated clients, such as systematic hedge funds, but also with other banks, says Dr Ralf Donner, Head of Marquee Execution Solutions at Goldman Sachs. “As one of the leading algo providers we are active in marketing our product to the spot desks of smaller banks, who perhaps do not have the same means or the same variety of tools for execution. Franchise Matching is an  exciting development and has proven to be very successful with these desks, in providing something that is competitive compared to direct market access,” Donner says.

He adds that it is however worth noting that there will be times where it may not be suitable, such as Goldman Sachs having positioned in the same direction as the client who is attempting to trade. This is why it is important to have control and for the Franchise Matching functionality to not be set in stone, Donner explains. “It is not necessary to suspend or cancel the algo if Franchise Matching is too slow. Clients can easily modify their access to different categories of liquidity, selecting additional liquidity or even unselecting Franchise Matching if it isn’t effective, and choose other forms of liquidity instead,” he says. “Another reason why it has done so well is that we also have internal demand for this tool, with our own voice traders now actively using it for their risk management purposes.” 

Innovation and development

The analytics results from third-party TCA are also making the offering more attractive to clients, Donner adds. Clients have started looking at this data more closely and increasingly it is not just about overall all-in performance, he explains. “The TCA may include certain specifics, such as performance by individual venue,” Donner says. “Clients are now starting to ask questions about the venues that they’re accessing when they trade with our algos. It has become increasingly important to demonstrate value and, when it comes to internalisation for instance, to have made that as high quality as possible. This is what we have done with Franchise Matching.”

Differentiation between algo providers is also going to be a lot more data driven than it has been in past, says Mesick. Customers now have more access to data from external providers as well as bank TCA, he explains. “How they interpret and use that data to make decisions about where to go with what flow is also becoming more sophisticated. Our relationship with customers is very important but these conversations around algo performance are becoming much more data driven than in the past,” says Mesick. “Differentiating between the various algo providers is going to come down to performance. If your algos are performing, you’re far more likely to continue executing with those customers, while the providers who have issues accessing liquidity are going to be more easily identified.” 

Another effect of better analytics data is that clients now have a better understanding of how their execution impacts the markets, notes Mesick. He adds: “This is one of the main reasons why internalisation has been so key, especially in our passive algos suite. We have increased our internalisation rate of passive algos by around 4x in the past year, which is substantial.  We are already seeing the benefits of that change from a TCA perspective. Ultimately, it is now one of our biggest differentiators as an algo provider.” 

Improving the user experience

A further differentiator which stands out for clients is in how well FX algos can be integrated into their existing workflows, Mesick says. “Clients are now more comfortable and familiar with the algos, so now they are increasingly looking at how the algos fit into their workflow, how they can be more efficient and do the same amount with less. Seemless workflow integration is now another important factor for algo clients, especially in a cost constrained environment as rates are obviously in a much different place than they were a year ago,” he adds. Kataria agrees, adding that this was also reflected in the findings of the client survey results. When asked what the most important consideration was when reviewing the performance of dealer algos, internalisation was the number one response, at 33%. “Just behind internalisation however, the second most important factor was configuration of algo at an order level,” says Kataria. “We differentiate by performance as well as making sure we offer what our clients need, such as workflow facilitation.”

The role of analytics will also be increasing based on observing how a client trades in order to then suggest ways to improve their execution, adds Nuti. He explains that this ties in to the ability to customise orders. “Our remit is to optimise capturing that execution in a reasonable timeframe. But some clients may have very short term alpha, where others might have very long term alpha. So we might have more time to execute, which is another way of customising the execution,” says Nuti. “We have had successful engagements with clients who trade very small sizes, to very large sizes, all through FX algos, and the performance is very sustainable. Optimising workflows streamlines the process for clients, so instead of spending time in trying to manage various LPs or multiple connections, they just have one or two. These are all hidden but real costs, which we can help to minimise.”

For algos, Nuti says that Deutsche Bank currently offers some of the widest coverage on the market. “We offer deliverables in both G10 and EM pairs, while also covering LatAm and Asian NDFs alongside precious metals.” Hedge funds are also expected to increasingly adopt algo execution, according to Nuti, while the size at which they start using them will change as well. “The average algo client might currently expect to use an algo to execute an order size of at least 20-30 million, but the size at which they will start to consider using an algo will likely get much smaller,” he adds. 

Clients are becoming increasingly sophisticated and demanding better quality assessment of the algos that they are employing

Areas of new growth  

A key form of internalisation at Barclays is the ability to use an FX algo to access the franchise liquidity, notes Mesick. “Our Barx Gator algo suite in particular allow us to show interest to our franchise directly. Being able to tap into that franchise has really vaulted our internalisation rates and is one of the main ways we think about internalisation.”  

Bringing together the concept of the evolving liquidity landscape with customer demand for NDF algos also makes sense as the two are related, according to Mesick. The maturation of the underlying liquidity in the different electronic venues for various currencies is a lot more varied than the deliverable space, he explains. “We already support NDF algos, but as key markets continue to mature and electronify further, we will be able to offer increasingly advanced algos to trade.”

Over the past decade, one question which continues to be raised by FX algo clients is what volumes are being traded in a particular currency pair, according to Donner. “This is still a difficult question to answer. One development is that we are currently helping to contribute to a third-party vwap benchmark in conjunction with one of the third-party TCA providers,” he says. “They have a client request for a vwap benchmark from a variety of different banks and we are contributing to that. As contributors, we then also get to see the final benchmark. This is the one of the first concerted attempts to produce a vwap benchmark for FX.” 

In house, Goldman Sachs is also currently building out a pre-trade toolkit utilising live market data, Donner adds. “The fundamental concept is similar in design to a tool that we had launched on marquee previously for options structuring which we call the Visual Structuring tool. This is essentially an analogous version of that, but used to make the decision between trading with an algo or trading using a risk price. Equally, it tries to show all aspects of market liquidity as well as possible algo outcomes versus alternatives on a single chart. This is going to be a very powerful tool which we can share with our clients in the very near future.” 

The end of AI in algos?

When it comes to the future utilisation of AI in FX algos, Donner re-asserts the position that AI has no place in this very limited set of markets. “AI can be useful for algo trading in other markets with far more instruments to be traded, such as futures. But in FX, it’s just not worth the risk of a very unusual, unexpected algo outcome in even a tiny fraction of cases.” Nuti adds that in the case of AI, the biggest issue is being able to use it in a highly regulated industry such as FX. “You go from A to B, but you cannot explain how you got there. That is difficult to explain to a regulator and not a place most institutions want to be in,” he says. “Yet machine learning, reinforced learning, econometrics, things that are much more deterministic in their outcomes, are proving to be a much better area of focus. The algo market is going to continue concentrating on this area, rather than AI.” 

According to Barclays’ client survey findings, clients are also increasingly allocating their algorithmic execution based upon historical performance. “Clients have become more sophisticated in their use of data. When we asked them how they allocate their flow to dealers, the largest response at 29% said this was based on historical algo performance,” says Kataria. “As data analytics comes more into play, we will see clients building more of algo wheels. Data driven decision making is key in algo trading.” Automation will also prove key as the use of algo wheels increases, with more clients investing in data and analytics behind these tools, says Mesick. “This will automate algo selection and provide a better view as to where they need to be routing their orders,” he adds. 

Another question which continues to be raised by FX algo clients is what volumes are being traded in a particular currency pair

Towards maturity

Generally, the FX algo market is now fairly robust, says Kataria, with some 37% of clients taking part in the survey having said that they now leverage dealer algos. “This follows the general industry trend of the algo market now being fairly mature,” he adds. “At Barclays we have a very mature product given that we were one of the first banks to deliver algos. So at this point, for us, it’s about a lot of small tweaks that have large back end impacts. And that’s been our biggest focus for 2023 and will continue to be a focus for 2024.  Internalisation is one of the things we have been talking about the most as it’s had the biggest impact on our client performance and we think it will continue to have that impact,” he concludes.