Algo trading has witnessed an upward trend over the past eight years in FX. Despite algo adoption trailing the likes of equities, there is no doubt that the use of automated execution methods is on the rise in FX. First referred to as expert advisors or EAs, algos entered the FX world 10 years ago according to a Bank of International Settlements (BIS) study.
These days FX spot witnesses the highest algo usage and volumes. According to the BIS study, algos account for 10-20% of trading volumes, amounting to approximately USD 200-400 billion in daily turnover. Algos have evolved over the 10 years they’ve been present. Initial iterations sliced larger orders into smaller chunks and executed them piecemeal.
Currently, algos react to market conditions in real-time and help traders adapt their strategies to changing liquidity profiles. Despite these advances, traders undoubtedly seek improvements and “nice-to-have” features. What are some of these, and can algo service providers rise to the challenge?
Factors Influencing Algo Usage
Andreas Anschperger, Director, Senior FX Trader, Fixed income and FX Trading at Allianz Global Investors, is clear about the enhancements he would like to see. “Easy access to the entire algo suite, including the guiding analytics during pre-trade / at-trade/post-trade are some enhancements I would love to see,” he says. “Insight like this will allow us to create a more efficient and transparent STP process.”
Richard Turner, Senior Trader, Currency Solutions at Insight Investment echoes this view. “Transparency is a key factor when executing algorithmically,” he says. ” Service providers should offer algos that allow the user to adapt their execution style over the period of execution whilst maintaining transparency. Transparency should include all aspects of the child fills and any actions taken by the trader over the period of execution.”
A recent study conducted by Coalition Greenwich highlighted that transaction analytics and algo usage are joined at the hip. Algo usage prompts greater TCA usage, and increased TCA builds trader confidence in algo execution. These findings are in line with Anschperger’s thoughts. He also highlights more competitive fee offerings as something that would be nice to have.
However, the gap in current TCA offerings is a stumbling block. Jan Grindrod, Global Head of FX Trading at Invesco, points out this issue. “There are large differences between providers’ offerings when we look at pre and post-trade TCA,” he says. “It would be nice to see that gap closed. Pre and live TCA give traders the ability to set expectations regarding how algos will perform and monitor their executions in real-time.”
The promise of receiving better prices than risk transfer prices is a driving factor behind algo usage. Thanks to smoothing large orders in illiquid pairs, algos can potentially manage the varied factors behind trade execution easier. Some of these factors include the urgency and nature of the order, the order’s size, time of the day, market dynamic, the spread, and the liquidity profile of the pair. Algos can give traders an edge in managing these factors efficiently.
However, this ability has to be balanced with algo providers offering traders insights into trade analytics at every stage. Some algos offer traders the ability to increase or decrease their participation at certain levels. Switching strategies in-flight and evaluating market conditions in terms of execution costs are also features in which traders find immense value.
Grindrod highlights how Invesco’s use of algos has changed over the years, thanks to increasing sophistication. “In our first use cases, we primarily used time-based FX algos to execute large orders in liquid currency pairs,” he says. “Since then, legacy FX algos have become more sophisticated; more FX algo types have been developed; and pre, live, and post TCA tools have become available.”
“As a result, algos have become a more standard tool in the toolbox, and we’ve become more dynamic in the ways we use them.” He also points out that greater algo usage has increased the number of currency pairs traded and Invesco’s order size range.
Turner explains how technological progress has changed the way he uses algos. “We were one of the very early adopters of FX Algos evolving from “click and deal” electronic execution,” he says. “In the early days of algos, algo behaviors were simple and akin to that of a VWAP or TWAP. Since then, algo providers have created more sophisticated algos that will behave in different ways to match the user’s varied execution objectives. The more sophisticated algos now enable the user to change parameters as the algo is trading. One area where we have seen significant development is pre/post-trade TCA. Several algo providers now give the user tools to decide the optimum algo strategy to match their execution.”
Reducing market impact is a significant consideration when executing trades. Currently, trades that take place over several days or weeks and involve different asset classes lend themselves very well to algo execution. Analytics packaged with algo services help traders minimize costs and improve entry and exit levels.
As Turner puts it, “Ultimately, we do NOT want to signal to the market that we are trading. We continually analyze the TCA data to ensure that this is the case.”
“Accessibility to a liquidity provider’s algos across currency pairs is key,” says Invesco’s Grindrod. “If a particular algo works for pairs that trade in the primary market, but cannot be used to trade crosses or NDFs, it increases the probability of a user having a bad experience and not coming back when an order is rejected.”
Thus, improving analytics and visibility is a perennial request from the buy-side. The better the analytics are, the higher algo adoption will likely be. Data captured from these reports also inform future execution practices, driving significant change over time.
The Changing FX Landscape and Algo Usage
Has the regulatory landscape had an impact on algo usage at buy-side firms? On the surface, Phases 5 and 6 of UMR have put trading costs and efficiency under the spotlight. The cost of a trading seat is now under scrutiny, and in this environment, the efficiency gains that algos promise might spur increased adoption.
However, Allianz’s Anschperger says that this hasn’t been the case. “Not really,” he responds when queried. “The presence (or absence) of liquidity is a stronger driver,” he says. Grindrod echoes these thoughts by agreeing that UMR specifically hasn’t affected Invesco’s use of algorithms. Turner gives a more guarded opinion by noting “there haven’t been major changes…yet.”
Despite changing trading economics leaving algo adoption relatively unaffected, there’s no doubt that FX market stakeholders are increasingly accounting for algo activity. The GFXC recently updated its Global Code and published guidance regarding pre-hedging and the Last Look period. The Code standardized algo disclosure and TCA templates, reflecting the increased interest in algo adoption.
Anschperger believes that while these changes are needed, they don’t impact algo adoption to a large extent. “Taking “no last look” liquidity into consideration might be an impact from these regulations, but generally, our algo adoption won’t increase or decrease.”
Grindrod points out that Invesco has been using algos for many years now. “While we welcome the changes to the GFXC’s code, the changes in the code by itself will likely not result in us executing more orders via algo,” he says.
Best practice guidelines from the GFXC help firms that provide liquidity but are also price takers. In turn, this gives firms the freedom to test and explore diverse strategies, aided by algos. Ultimately, the lesson here is that greater trade efficiency and price discovery are driving algo adoption higher.
“We would like to see algo providers produce standardized TCA, utilizing the template promoted by the FX Global Code,” says Turner. “We would also like to see disclosures, as set out in the FX Global Code, from the algo providers pertaining to the behavior of their underlying liquidity (last look policy & hold times.) This will enable the user to be as informed as possible about the behavior of the selected algo when executing in the market.”
Greater data availability has changed buy-side expectations to a large extent. “Access to more granular data about market dynamics and liquidity has resulted in pre and live TCA models to be built, which has helped the buy-side make more informed decisions about how to route FX orders,” explains Grindrod.
“It’s also allowed the buy-side to become more systematic about how FX orders are routed,” he continues. “Having more market data readily available has helped us become comfortable using more sophisticated FX algos, as we are better able to set expectations regarding how algos will perform and monitor their executions in real-time.”
Product Expansion and Changing Work Conditions
No discussion of the markets these days is complete without acknowledging the effects of the COVID pandemic. “Before the COVID-19 pandemic,” says Grindrod, “I think there was a perception that while the FX spot market had become heavily electronic, market participants would reverse course and pick up the phone to execute orders during volatility spikes.”
However, Invesco’s execution dataset combined with data from sell-side highlights increased algo usage in the second half of 2020. Restricted to working from less-than-ideal conditions, firms began relying on algos to smooth the transition. The message was clear: FX algos aren’t tools for just good times. They’re invaluable during rough weather too.
Algo usage helped firms combat increased volatility in the markets. In turn, these conditions helped traders familiarize themselves with algo usage. Anschperger notes, “We definitely witnessed higher algo usage during those volatile market environments. Higher spreads in those times caused more traders to lean on algos to smooth execution.”
Turner points to similar lessons learned during the initial days of the pandemic. “Trading electronically (via algos) enables for a great deal of transparency during periods of extreme volatility,” he says. “It allows you to quickly access liquidity when it is available. The other key element is that you can evidence the best execution and other regulatory requirements very easily.”
Grindrod is quick to point out that greater algo adoption doesn’t mean the elimination of voice trading. “The pandemic was a special situation, one where volatility was high, volumes were high, and risk prices were wide,” he says. “In this situation, traders were willing to take on a bit of market risk to potentially save on bid/ask spread.”
The FX market is highly fragmented, with algo usage in NDFs and options low. However, algo adoption is set to increase in these markets. Anschperger believes NDFs will witness greater algo usage moving forward. “Less liquid EM markets will also witness more adoption,” he says.
However, NDF adoption faces stiff challenges, says Grindrod. “Some liquidity providers have started to expand their algo suites to NDF pairs, but there is still much work to be done on this front,” he says. “Even if the buy-side gets comfortable routing NDF orders to algos, the trading workflow up through the point of order entry is more manual and onerous for many buy-side clients using multi-dealer platforms for execution. Not all platforms allow for or can handle NDF algo workflows.”
The result is traders relaying orders to sales desks and avoiding the use of EMS’ via FIX. Clearly, product enhancements to support greater NDF execution are in order. Grindrod also highlights a few more enhancements.
“Buy-side traders need to keep track of which liquidity providers’ algo suites support NDFs, which algos within the suites can be used for NDFs, and which NDF pairs are supported,” he says.
Insight’s Turner highlights specific functionality issues that would make or break his choice to use algos for NDF execution. “We trade to “broken dates” and algos typically trade to the most liquid date such as 1month or BMF dates,” he explains. “We have a couple of providers who support broken date execution, and we have traded in this manner with them. The execution outcomes have been very good, and we will continue to evolve in this space.” However, he notes that service provider consistency is an issue.
“We need our other providers to support this construction so that all child orders can be assessed, rather than having huge swap risk over the period of execution,” he says. “Transparency is key, and if all providers offered broken date functionality, we would increase our NDF Algo volumes.”
Rising Use Creating Varied Needs
As algo usage increases, buy-side trader demands are likely to increase too. No doubt, increasing comfort with algo execution and greater use of TCA are driving algo demand higher. It remains to be seen whether service providers can meet trader needs.
Overall, the mood is optimistic. “I think the FX Global Code has been a big influence in “doing things right” when trading,” Turner says. “The transparency that algo trading brings has allowed traders to hone skills in constantly improving their outcomes in an ever-changing world. Add to this, rules-based automated trading with appropriate compliance systems, and the trader’s job really does become interesting. It allows us to focus on the bigger risks in our businesses.”