Liquidity in algorithmic FX trading: Analysing the key dynamics and strategies at play

The uniquely fragmented nature of FX liquidity has arguably been a key driver behind the growth in adoption of FX algos, particularly among buyside firms. Leading FX providers are also aware of the impact good liquidity management can have on execution outcomes, with many now actively developing additional tools and services to help meet this demand. Nicola Tavendale investigates.

Industry Views
by Nicola Tavendale

Liquidity has always been a critical factor when it comes to FX algo performance, but it is even more of a differentiator now that the majority of algo providers offer a wide range of algos and related functionalities, says Carolina Trujillo, Head of e-FX distribution at SEB. She explains that this differentiation is less about the features and more about the logic involved, which includes liquidity and good liquidity management. “What at first appears to be similar liquidity can in fact vary a lot, depending for example on which venue it is, the fill ratio of that venue, whether it is firm or non-firm liquidity etc,” Trujillo adds. “The algo provider needs to constantly analyse all venues, selecting and prioritising the liquidity sources it offers to the client in order to ensure they can achieve the best execution possible.”

Another key point that will impact liquidity is how the provider behaves on the venue, as this will directly translate into the quality of the liquidity that the provider will be able to access, she says. “As a result, two providers accessing the exact same venues can have access to very different liquidity,” she adds. According to Trujillo, an extension of this is also how much a provider will police and dialogue with the venue so that the liquidity that they have access to is of good quality, such as by removing participants with a too low fill ratio, for example.

“One of the great benefits of working with professional clients is that the clients are engaged and want to understand the background,” Trujillo adds. “This results in very interesting discussions that also can be the start of new ideas for the algo providers. It is beneficial for buyside firms to understand about the underlying liquidity so that they understand why a venue was selected above another one, or why a venue is not part of the pool the liquidity provider is going to use, and they can feel more comfortable with those choices.” In addition, she notes that once buyside firms are clear in their mind of what they need from a liquidity standpoint this will also help to look at how they can best achieve that when selecting the right algo. “Education to clients is still a very important responsibility for algo providers,” says Trujillo. “The algo disclosure from the FX Global Code is also very thorough and has definitely helped buyside firms to compare and better understand how the pools are created and reviewed. The questions we get from the clients that go through those disclosures are a lot more specific and informed than those that do not look at the disclosure,” she says.

Utilising the data

“The algo disclosure from the FX Global Code is also very thorough and has definitely helped buyside firms to compare and better understand how the pools are created and reviewed.”
Carolina Trujillo

Harshad Hariharan, FX algo product manager at JP Morgan, adds that in such a diverse and fundamentally complex market as FX, liquidity impacts almost all facets of trading. “When it comes to FX algos, performance is largely dependent on the quality of liquidity sources available for execution. Executing externally may have a different impact post execution when compared to an internal fill,” Hariharan says. For example, he notes that based on historical JP Morgan analysis, fills against external liquidity may impact the price in the long run – potentially due to information dissemination – while the price impact of internal fills trend to flat. “Internalisation is therefore a critical performance metric for clients that are looking to minimise market impact on the back of their executions,” he explains.

The external liquidity landscape in FX is also constantly evolving and new developments in how clients can access liquidity, such as full amount trading, anonymous/dark pools etc, are gaining in popularity, Hariharan says. “For instance, a key consideration when selecting an algorithm is minimising market impact and strategies like full amount trading can help achieve this,” he adds. In addition, Hariharan notes that regulatory developments, such as MiFID II in Europe, have also added to the drive for greater pushed transparency around execution quality and this in turn has led to an increased uptake in the use of TCA reports by the buyside to evaluate their FX algo performance.
He adds that clients can utilise such TCA reports to identify useful key factors, such as which liquidity sources are providing optimal algo execution while minimising market impact, or to recognise the times of trading where liquidity is higher or lower than usual. “Clients can also use TCA to help understand if the FX algo strategy they have chosen performs best with the liquidity sources available for execution. In addition, JP Morgan offers TCA during all phases of algo execution to ensure complete transparency” says Hariharan.

Evaluating market impact

“Internalisation is a critical performance metric for clients that are looking to minimise market impact on the back of their executions,”
Harshad Hariharan

Furthermore, while interaction with immediately availability liquidity has always been the central theme for algo performance, the increasingly complex nature of the FX market means that it now also presents one of the most significant challenges, says Peter Nuttall, Head of FX algo product, MUFG Bank. “This can only be met by delegating liquidity access to a sophisticated smart order router (SOR) that has either been designed specifically for FX liquidity access, or highly adapted to it,” he adds. “Of particular importance is access to internal liquidity. Our SOR not only provides our FX algo clients with the ability to aggress our internal liquidity alongside external liquidity but, much more significantly, facilitates franchise access to our unique and globally distributed streaming price client network. Although this presents an onerous technology challenge, it means our FX algos can simultaneously interact passively with both internal and external liquidity – to the tremendous benefit of our FX algo clients.”

According to Nuttall, due to the widespread adoption of FX algos and TCA reporting in recent years, buyside firms have also become increasingly aware of the idiosyncrasies of FX liquidity, rather than it being obscured from view behind a risk transfer price. He adds that this insight leads to a nuanced understanding of both external and internal FX liquidity sources which he argues has allowed buyside firms to better conceptualise their preferred style of execution and express their execution urgency. “Many providers offer customisation of their FX algos having acquired this deeper appreciation of FX liquidity,” says Nuttall. “More and more clients are interested in customisation of how their algos access liquidity in order to achieve their trading objectives. TCA reporting plays a key role in this process because it enables continual discussion on performance and the part played by access to the various sources of liquidity.”

Liquidity management is also critically important when looking at the overall cost of execution and for its role in helping the algo provider or client to understand the market impact when executing that flow, adds Vittorio Nuti, Global Head FX Algos at Deutsche Bank. He explains that FX algos are a product and, as such, buyside firms benefit from understanding all facets of the service being offered, especially considering the impact that the underlying liquidity involved can have on lowering their cost of execution. Yet he warns this will also vary depending on how the client has opted to execute their order and why. “Sometimes a client will give the algo provider their full order to execute in the best way possible within a set timeframe and this may cause some market impact. Then at other times, the client may prefer a very passive execution and to use the curated liquidity offered by the provider. But in either case, what you don’t want is for market impact to accumulate and result in a higher overall cost of execution than was necessary,” says Nuti.

Learn and adapt

“More and more clients are interested in customisation of how their algos access liquidity in order to achieve their trading objectives.”
Peter Nuttall

As an algo provider, Nuti says that discussing the liquidity selection processes with clients directly and providing real time TCA is key. However, he notes it is also very important to understand that Deutsche Bank is able to evaluate these impacts across multiple executions and using the globality of its trades, rather than a single order. He adds: “This is because ultimately the variance is going to be quite large, depending on any given child order or any given parent order. So the best way to assess this to a useful degree is to try and keep the set of observations as high as possible.” Taking a more data driven approach to liquidity management is therefore essential, he argues, as it can be too easy for clients to be overly focused on one bad trade. “Following a poor execution, clients should adopt a quantitative approach, taking a look at the data and then coming to a decision,” Nuti says.

Market impact should also not be the main consideration as all FX algo executions will result in some level of market impact, Nuti adds. “If a client wants to place an order that is visible during the child order placement, then there will be some market impact – but it will also give them a chance to get that trade and potentially earn some bid offer spread. It is all about opportunity cost, clients need to really weigh up the availability of liquidity, the average spread that they need to cross and so forth, when they are assessing child order placement. Avoiding market impact shouldn’t be the only aim”, Nuti says. Trujillo agrees that there are so many different parameters to verify and regularly update in terms of liquidity management that it can only be achieved with a quantitative and systematic approach. In turn, she says that a data driven approach also enables the algo to learn from the historical information in order to improve and adjust parameters for the best outcome.

According to Trujillo, liquidity access and liquidity expertise has become a key differentiator. Since many competing algos are quite similar, she argues that the smartest access to liquidity – either via internal pools that are superior, or via a smart set of rules to determine where the best execution can happen – is what will make the difference. “One of the ways to leverage liquidity expertise is by the quality of the internal pool that the buyside firms are able to access. This is one of the main advantages buyside firms will get in using SEB algos due to our Scandi focus and unique Scandi franchise, she adds.

Looking at liquidity sources

“Taking a more data driven approach to liquidity management is essential.”
Vittorio Nuti

Trujillo says she also welcomes the recent creation of more bespoke liquidity pools reserved to FX Global Code compliant participants, adding that as more venues go that way, it will be easier for algo providers to select from a subset of venues, adding: “This is something that SEB supports and that all participants should strive for.” On top of that “Changes in the market, changes in the venues, new rules, newcomers into a venue that can disrupt it, the constant evolution of the FX market makes it absolutely necessary to periodically review and adjust the liquidity sources you are using or adjust the logic into your own pools,” she explains. “The algo business is typical of the FX business where you have to tweak a multitude of parameters, rethink, re-invest and adapt to changes constantly in order to stay relevant. It is also what makes it so interesting and exciting.”

Hariharan says that historical analysis of liquidity provision can reveal useful insights for the algo provider, such as the observation made by the bank that faster round trip times give liquidity providers less time to decide whether to fill or reject based on market conditions. As a result, market impact can be significantly reduced by minimising response times, he explains, adding that this is one of the primary reasons why JP Morgan repeatedly tracks this metric with its ECNs. “We have worked extensively over the past few years with ECNs to improve response times and have seen our efforts get returns with significant improvements in algo execution,” adds Hariharan. In addition, JP Morgan has also developed an in-house customisation framework that helps cater to various client requests, including customised liquidity, as well as supporting bespoke liquidity pool requests from clients.

“FX Code Only is an example of a bespoke liquidity pool offered by JP Morgan to allow clients to access our in house liquidity, aggregated with others that have also attested to the FX Global Code,” he adds. “Our observation so far, however, is that the majority of clients continue to rely more on the algo provider’s assessment of best available liquidity sources.” According to Hariharan, more of a critical differentiator is how an algo provider manages its external firm/non-firm liquidity pool, but warns that liquidity management is an effort that requires a lot of up-to date data and resources. “In addition, internalisation significantly reduces market impact – which is the execution objective of most clients – and hence it should be one of the most important factors to look at when reviewing FX algo performance,” he says. “We continue to invest and innovate in this area. This has translated into our algos managing to internalise the bulk of our execution volumes over time, with our latest being 50% flow internalised in H1 2022.”

Furthermore, due to the diverse landscape of FX liquidity sources and the need for cutting-edge SOR technology, emphasis then falls on making best use of the large amounts data in a real-time and systematic manner to optimise execution performance, Nuttall says. He explains that this is achieved in part in the SOR by maintaining a statistical picture of where the highest quality liquidity resides at any given point in time and how it is expected to vary in response to trading. In addition, Nuttall adds that from a longer time horizon perspective (and despite the fact that FX lacks the consolidated tape of listed markets) an increasing number of FX venues and settlement firms are providing enriched and alternative sources of market data. He believes that when this data is suitably processed by machine learning techniques, it can then provide a holistic view of intraday FX liquidity that he says allows, for example, execution horizons to be optimised and alpha signals to be generated that attempt to anticipate market direction.

TCA reporting enables continual discussion on performance and the part played by access to the various sources of liquidity

Innovations and further evolution

“Providing franchise access to MUFG’s streaming price client network is a significant value add for our FX algo clients, as it helps reduces market impact further than would otherwise be possible.” says Nuttall. However, the overall network is constituted from individual client liquidity pools with differing information leakage profiles, he explains. According to Nuttall, the curation process essentially calibrates the urgency of execution to the information leakage of each client pool. In that case, when execution urgency is low, access is only provided to low information leakage pools but where it takes longer to attract opposing interest, whereas when execution urgency is high, access is provided to pools with higher information leakage but where opposing interest is attracted almost immediately, he explains.

At a minimum, he says that MUFG’s FX algo clients can choose from internal, external or hybrid (both external and internal) liquidity options and, within each of these, liquidity access can be customised. However, he notes that many firms who provide an FX algo service as well as providing streaming prices typically operate the two enterprises in a siloed manner. “Although there is a clear regulatory need for segregation, there are opportunities to create synergies that benefit buyside clients and provider alike, whilst remaining pursuant to regulatory obligations,” Nuttall adds. “MUFG allows the FX algo and market making businesses to coalesce whilst conforming to the FX Global Code; not least to facilitate franchise access, but more broadly, buyside clients benefit from the combined expertise in liquidity management and shared investment in cutting-edge research and technology for optimal access to FX liquidity.”

Overall, having a solid liquidity that is varied in nature and is there at all times should be considered a far more important thing than if the execution results in a tiny amount of market impact, Nuti argues. Due to the Deutsche Bank’s quantitative approach, it has also gone down the route of developing much more bespoke liquidity with its bilateral relationships, adds Nuti, along with curating the liquidity interaction with the bank’s franchise. “In addition, the development of Pro Liquidity allows us to show our algo interest to clients that have genuine willingness to trade and cross spread,” he says. “We believe is this a unique feature for this market. We’ve seen great results with it already, with north of 20% of our flow of algo execution now going through that venue, which is far more than any mid book venue execution that we’re aware of.” Due to FX being an OTC market, it also has a vast array of different styles of liquidity, different types of liquidity, different spreads etc, explains Nuti. “As a result, curating that liquidity is the initial quantum of providing a good service to our clients. It is then more about finessing that execution and making sure that the algorithms themselves aren’t the cause of unwanted market impact and so forth. But that initial liquidity sourcing is very important as an algo provider,” he concludes.

Related Posts