A live survey of the 773 attendees who joined the conference online across the two days found that 67% of those responding considered improved FX algo products to be the most important development to help their firms source new liquidity over the next 12 to 18 months (Figure 1). A further question took a closer at what had been the drivers for FX algo use, where the various events of 2020, the US election, volatility spikes etc, had prompted some 53% of industry participants to increase their adoption of algo trading, while only 15% believed it had remained the same and 13% were less reliant on algo trading than before (Figure 2).
Over the coming year to 18 months, 30% of respondents also expected the use of AI and ML to be increasingly used in execution algos. When it comes to TCA, 59% reported that their firm was currently using some kind of TCA solution, with 16% stating that this was mainly used to aid their algo selection (Figure 3).
Panel discussion
These themes continued in an Automation Panel Discussion, where panellist Thomas Gentile, Director of Trading at Crabel Capital Management explained that as a buyside firm, they had been using electronic algorithmic execution methodologies in the FX space for many years now. He added that firms might even be considered a little bit remiss if they weren’t taking advantage of market movements or not impacting the market as much as they possibly could by not operating on a 24 hour basis. “We develop all of our own algos in house and we trade on our own developed infrastructure. We’ve been trading NDFs electronically for quite some time now and we’re happy with the outcome,” Gentile added.
Whilst Brendan McMurtray, FX Electronic Trading & Market Structure Analyst, T. Rowe Price said that his firm’s use of FX algorithms grew by almost 5% in March and April of 2020. As spreads have since tightened and sell side desks have started to find their footing again, he adds that FX algo usage has subsided but has still settled at levels higher than those seen pre-crisis. “Now, when we think about this, relative to automation, we don’t currently deploy algos within our automated trading system. But if we had, we may have considered pulling back on their usage in high volatility and high spread regimes, given the concerns around data availability,” he added. “It is critical to have a system that can be adjusted to reflect different market regimes and you have to have the data necessary to make statistically significant claims about your decision, otherwise you’re just picking an algo out of a hat.”
Figure 1: Sourcing of Liquidity
Source: Aite Group: Live Survey of TradeTech FX USA attendees from February 9, 2021 to February 10, 2021
Figure 2: FX Algo Drivers
Source: Aite Group: Live Survey of TradeTech FX USA attendees from February 9, 2021 to February 10, 2021
Figure 3: Motivation for Using TCA
Source: Aite Group: Live Survey of TradeTech FX USA attendees from February 9, 2021 to February 10, 2021
Buyside concerns
In addition, McMurtray notes that there have been three primary concerns from the buy side which have stood in the way of greater adoption of NDF algos, particularly for traditional asset managers. “This includes the typically higher fees associated with these offerings from the banks and handcuffing ourselves to our algo provider to roll our one month NDF algo execution to broken dates,” he says. “In terms of rolling NDF algos from standard to broken dates, which for us, given that we trade broken dates frequently, this is a particular concern.” He adds that several providers have come up with innovative offerings here, such as continuous swap hedging throughout algo execution and agency swaps liquidity provision, but in terms of access, there have been limitations from some of the most widely used multi dealer platforms in their ability to support these offerings. “While the concerns are primarily regulatory in nature, this is definitely something that needs to be resolved before NDF algos can truly take off amongst the traditional asset manager community,” McMurtray added.
Platform perspectives
Matt O’Hara, CEO of 360T Americas, also reported an increase in algo usage during the period of market dislocation last year, noting that existing users of algos deployed these tools more frequently across 360T’s platforms while some clients actually began using them for the first time as spreads widened out and it became harder for them to effectively execute full amount trades. He noted that trading non-Spot FX products in a manual fashion was particularly challenging during this time, but added that firms with access to the right data and automated trading solutions were able to overcome this. “High-quality data combined with automated trading gave buy-side firms that were using these tools the ability to execute much closer to the mid than those who were trading electronically but executing manually. That was very interesting because I don’t think many people expected that outcome,” he said.
Meanwhile even though FXSpotStream’s focus is very much on automation, Tom San Pietro, CTO, FXSpotStream, explains that the trading strategies that work best are those that support building relationships as well and it is those human interactions that make trading sustainable. He noted that their upcoming new algo product (see page 3) will follow the same successful model, enabling access to the FX algo suites of the FXSpotStream LPs through a single network connection. “We’re trying to take things in the FX algo space to the next level. We’re going to introduce an API that supports over 70 algos and hundreds of parameters. It just seems odd to us that at the end of all the TCA and data analysis that has to be done to decide which algos you want to use, you still need a human to go sit in front of a screen and punch in all the parameters and pick your algos and send things off. We think it’s a logical progression to move away from that.” Yet he says from their experience of getting this product rolled out that human relationships and interactions are even more necessary, but that automation and efficiency just allows the conversations that people have be more based on data analysis and to add value.
