This year’s conference featured a panel dedicated to exploring the use of algos, both in-house and third party, with high level discussions about the rising profile of algo execution and the significant role they have established in FX. Moderated by Stephane Malrait, Chairman, ACI FMA, the panel featured a mix of leading figures from both the buy and sell sides, namely: Ralf Donner, Head of FICC Execution Solutions, Goldman Sachs; Sana Horrich, Head of FX Desk, Banque de France; David Kalita, CEO, Quantitative Brokers and Baris Halitoglu, Trader at Nordea Asset Management.
Looking back on the panel and the event as a whole, Guild notes that the conversation on FX algos has matured significantly in recent years. Having previously featured on an algo panel at TradeTech in 2019, Guild explains that the main topic at that time was how to increase adoption of FX algos, whereas today there is a “much better understanding of algos and their use cases”. “The conversation has moved on to how buyside clients can better manage execution risk and how as an algo provider you can ensure they have the tools in place to help,” he adds. “On the panel we heard two interesting perspectives on how to achieve this.”
According to Guild, the risk premium on smaller trades means for many participants, the certainty of execution available from an RFQ will still outweigh the pricing benefit they might get from executing an algorithm. “But for larger transactions where there is that ability to absorb the execution risk and average it out over a large number of executions, there is now a good understanding in market that this is effectively the strong use case for algos,” he argues.
Demand for sophisticated toolsets
In relation, Chris Churchman, Head of Marquee at Goldman Sachs, had unveiled a new visual structuring tool at the Buy Side Innovation Day before the start of the conference. On the algo panel, Donner shared further insights about the innovation which aims to provide buyside traders with a ‘cockpit view’ of what’s going on in the market and what the predicted performance of the algo will be. Guild adds that effectively one of the biggest determinations for algo performance is “what everyone else is doing in the market at the time that the algos running”. “While this will always be more expectation than certainty, with the new tool this will be based on live information of what the algo is going to do against effectively the ability to pay that risk premium in terms of an RFQ,” he adds.
“So much of the algo conversation comes back to that truism – that the performance of the algo is dependent on what everyone else is doing in the market at the same time,” Guild says. “This activity is providing the liquidity that algos can then access and an algo is not worth using without the necessary liquidity.” The juxtaposition between in-house and third party algos was also highlighted by the panel, particularly around liquidity provision.
One of the advantages of using an algo that is provided by a major liquidity provider is the access to that provider’s internal liquidity pool, according to Guild. For example, Goldman Sachs’ offering is all about the relationship between the bank and its clients, with the algo product and liquidity provision being very much part of that approach, he notes, whereas with Quantitative Brokers, the buyside can access algos that effectively run independently of any particular liquidity provider. “Clients are then guaranteed against conflicts of interest or information leakage that could be associated with effectively using the algos from a from a particular liquidity provider,” Guild says.
Exploring the role of AI
“It comes down to a philosophical question of whether algo provision is something that is part of liquidity provision from a particular liquidity provider, or if it is something that is a little bit separate to that and best sits above the market,” adds Guild. “In terms of what we heard from the panel, there is not necessarily a right or wrong answer to that.” Essentially, this decision depends on the algo user and whether they want to be closely aligned to the liquidity provision that is part of your relationship with the liquidity provider, or if they prefer to access the market on a quasi-anonymous basis to access general market liquidity rather than a particular liquidity provider’s liquidity, he explains.
The panellists also shared their views on whether AI can improve algo performance, with the general consensus being that while this is the case, AI an also add complication and an element of uncertainty to the execution. “When the goal is putting tools in your client’s hands to help them manage the execution risk, that kind of uncertainty in terms of outcome and approach is not necessarily a positive thing,” Guild explains. “The potential benefits of AI and use of AI have to be balanced against the need for predictability and manageability.”
In the event in general, dialogue around the recent drivers of change in the FX markets appear to have also reached a more sophisticated level, including automation and the rise in algo usage, which was also in evidence at the event. The use of machine learning and Gen AI continues to be discussed as a potential disruptor, although according to Guild this is still an evolution that is in progress. “There is still some question as to what role Gen AI can play in in FX markets and also financial markets in general. That is not to say there are not potential uses in terms of optimisation of post trade processes and various other sort of functions around the outside of trading, but probably for FX execution there is less of a role for systems that are unpredictable and probabilistic in nature,” he concludes.
Pictures by Richard Hadley