Among the most interesting findings from the recent BIS report into FX execution algorithms and market functioning was that it was able to take stock of the performance of algos during the market volatility seen in March. With most providers reporting a doubling of algorithmic execution during that volatile period, some 30-50% more than the increase in FX turnover across all execution methods, the report notes that FX algos appear to have stood up well. With volatile trading conditions expected to become prevalent, what are the key elements which FX algo providers and their clients should be aware of to use FX algos not only effectively but also safely? Nicola Tavendale reports.
Since the arrival of FX algos, the liquidity landscape has been changing. Even though the size of the parent order is still the same the style of execution has changed, with a shift from people manually placing large orders to the algos placing the same order, but sliced into smaller orders which are worked over a period of time, says Asif Razaq, Global Head of FX Algorithmic Execution at BNP Paribas. “Now that algorithms are more heavily used, liquidity does not look the same and can look a lot thinner than it used to, but in terms of what’s physically traded, it’s probably the same amount but it’s just been presented in a different way,” Razaq adds. “The BIS wanted to understand the impact that FX algos are having on that new liquidity landscape and the additional challenges that might arise from that.”
According to Razaq, one possible challenge is the increased risk of ‘flash crash’ type scenarios. He explains that because the market may look thinner, there will be an increase in what look like volatile moves which could spark panic buying. This is one area of concern for the BIS. A further area of interest to the BIS was the controls that should be put in place to avoid the snowball effect which can lead to excessive price moves in the market. “As market practitioners, we believe that we are obliged to make sure that we adequate controls and circuit breakers in place,” says Razaq. “This is very important to ensure the continued safe functioning of the market.”
In keeping with this, a key area of focus at BNP Paribas is to train the algorithms to identify a broken market versus a market which is functioning again. Razaq explains: “The BIS report will also help provide new algo providers with guidance around what the key elements are that they should be thinking about when building execution algorithms, such as thinking circuit breakers, ways to minimise market impact on how to read that new liquidity landscape. As more clients are now using our technology and using our access to liquidity to take on that market risk, there is a duty for us as algo providers to ensure they have some cover from the volatility in the market and also to protect the market from any misuse of these algorithms.”
Reassessing the role of algos
Asif Razaq“As more clients are now using our technology and using our access to liquidity to take on that market risk, there is a duty for us as algo providers to ensure they have some cover from the volatility in the market..” |
The BIS report also noted that while the heightened volatility seen in March was not as pronounced as in previous crises, bid-ask spreads deteriorated, particularly for larger orders and to a greater extent in some instances than in other asset classes. Meanwhile, most providers reported more than a doubling of FX algo volumes relative to the average in usual conditions, with the most significant driver of increased execution algo volumes being seen as a relative deterioration in other methods of execution. In addition, volatility makes even the most heavily traded currency pairs more vulnerable and the impact of information leakage and large orders may have outsized impacts on price, says John Turney, Global Head of Foreign Exchange at Northern Trust. He believes that properly constructed algorithms have a better chance of minimising market impact, anonymising flow to reduce information leakage and so allowing the user to better avoid predatory trading activity.
“Ultimately, when trading in volatile markets, leaning on an algo strategy that focuses on reducing market impact by avoiding primary or secondary venues can be a sound approach,” he adds. “Access to liquidity and a laser focus on market impact are always tenets of a strong algo execution strategy. In volatile times another important component is flexibility: strategies that are designed to balance factors like liquidity and market impact have an advantage over those that are static in design in a fluid liquidity environment.”
Christian Gressel, Head of FX Algo Trading at UBS, adds that in volatile conditions it would help if trading desks had a better understanding about the duration of an order and the impact this can have. “After looking at the TWAP orders that we received during that time, we really found that clients in general were taking on far too much time risk,” he says. “While it is very, very good for trading desks to have the tool of choosing their own duration, there is a lot to be learned around picking the right amount of timing or duration that you want to give an algo. Ultimately, what clients really want to avoid is taking too much time risk, there’s always a balance between primary risk and market impact.”
Instead, Gressel believes clients should be aiming to find the sweet spot where they take the minimum amount of time risk paired with the minimum amount of market impact, avoiding taking on too much time risk if it is not really necessary. Educating clients about how the algos work should also be more of a priority, he warns, as although many clients are now active users of algos they may not be experts in how each particular algo works. He explains: “Clients need to fully understand the toolkit that each provider is giving them. A TWAP from UBS might look slightly different than a TWAP from another provider and the Float from UBS will look slightly different than from another provider.” Additionally, he believes active monitoring is essential. “As Q1 this year has shown us, none of us are experts on what’s going to happen in the next 10 minutes. Instead, we are good at assessing what is happening now, how this is impacting an execution and giving advice on possible alternative routes, it could even be as simple as speeding up or slowing down an execution. This active monitoring of orders has become a big value add that that algo providers can give to clients.”
Reproduced from BIS report: FX execution algorithms and market functioning
Completing large orders
John Turney“Ultimately, when trading in volatile markets, leaning on an algo strategy that focuses on reducing market impact by avoiding primary or secondary venues can be a sound approach,” |
Volatile markets can also lead to wider spreads which can make transferring risk more expensive. According to Andrew Cole, FX Algo Product Manager at J.P. Morgan, under these conditions limit based algo execution can work effectively since the spread savings (executed rate vs. top of book rate) can be far larger. “For some time, there has been a common narrative that algos ‘didn’t work as well’ in volatile markets,” he adds. “When these concerns first surfaced, we reviewed the risk transfer price savings from algo executions and looked at how they correlated with J.P. Morgan’s volatility index, but we did not find any strong correlation that indicated higher volatility deteriorated algorithmic performance.” Cole notes that client attitudes towards algo use in volatile trading conditions have evolved considerably since then, with many of J.P. Morgan’s record days for algo volumes often tallying with days of heightened volatility or the release of significant market news.
In particular, stressed and volatile markets can often play into the strengths of FX algo trading particularly for investors looking to complete large orders. The volatility spike observed in March might be seen as a good example of this. For example, Cole notes that the March volatility corresponded with a record month in fee paying order volumes for J.P. Morgan, with a total volume traded which was some 80% greater than the second highest month on record, June 2020. He adds: “March was also a record volumes month in terms of fee-paying orders as a percentage of spot, which showed that a higher proportion of volume than ever was conducted via an algo rather than click to trade.”
However, even more encouraging for J.P. Morgan’s algo business is the number of clients who continued trading orders even after the volatility associated with COVID-19 had subsided. Cole explains that among the reasons why clients who have switched to using J.P. Morgan’s FX algos still continue to use them is the range of execution tools available to them.
For example, it remains one of only a few banks which offers access to real-time charting on algo performance across multi-dealer and single dealer portals. J.P. Morgan’s aggregate TCA also went through a revamp this year and can now be configured to go out periodically to clients.
Focus on internalisation
Christian Gressel“As Q1 this year has shown us, none of us are experts on what’s going to happen in the next 10 minutes. Instead, we are good at assessing what is happening now and how this is impacting an execution” |
In order to operate a variety of strategies clients will need to better understand how they operate and when they should be used in various conditions. As a result, demand for sophisticated tools which help facilitate that is likely to increase. For example, BNP Paribas has invested in next generation components, such as its digital FX trading assistant ALiX. Razaq explains that if a client feels they are not sure what is happening with algorithm,
ALiX is now there to provide them with that full time commentary if needed. “That is also quite important to the safe functioning of the market, because as we see the use of FX algorithms increase, clients should also understand the inner work of the algos. By having real time tools available through ALiX, we’re positioning our clients to be more responsible when it comes to using algorithms,” Razaq says. “We’re now in this new world where a lot of our clients are interfacing with the market through algorithms, which means that we as market practitioners need to look at this new market landscape through a different lens as well.”
Market impact and access to liquidity are always primary factors that drive performance but in volatile conditions, larger orders can present even more challenges, warns Turney. He explains that in more subdued trading environments, buy side participants may look to use a single provider’s internalisation strategy to avoid signalling risk while accessing deep and high quality liquidity directly from the largest banks. “Yet while avoiding information leakage is prudent, its success can depend on whether a provider has the offsetting interest and enough liquidity to fill your order without taking on greater market risk by working the order too slowly,” he says. “At Northern Trust we focus our attention on sourcing the best available liquidity from each bank in our network directly, staying out of all lit primary and secondary venues. This lessens the reliance on any single liquidity provider becoming more defensive during volatile times, while still working to conceal any signalling risk to the market.”
The BIS report also noted that closing information gaps in terms of both expertise and data was central to ensuring greater transparency and a level playing field. “Market participants need to be able to assess the strength of their execution – pre-trade, in real time and post-trade,” the report continues. The technology underlying algo construction lends itself to granular detail, including execution arrival and risk transfer benchmarks, child order timestamps and venue analysis, Turney notes. He adds: “These details and analytics should flow into the reporting you receive from algo providers or third-party TCA solutions. Executing with an algo allows you to build a richer data set, giving you a better set of tools to measure and evolve your execution approach for future volatility events. Having accurate and detailed data can help you analyse how your approach worked in a volatile market and give you critical insight into how you need to evolve going forward.”
Reproduced from BIS report: FX execution algorithms and market functioning
Demanding more from providers
Andrew Cole“March was also a record volumes month in terms of fee-paying orders as a percentage of spot, which showed that a higher proportion of volume than ever was conducted via an algo rather than click to trade.” |
Overall, Gressel argues that in these volatile conditions, execution algorithms have passed their biggest test to date. He explains that when you consider the volatility seen in March, coupled with the disruption faced by every firm as trading desks and staff needed to be set up for working from home, then the market has been able to see that algos do work, even in the most difficult or adverse conditions. “Algos have worked very well and the performance of well-constructed algos has been good,” he adds. Even so, one of the biggest lessons he believes clients should takeaway from the experience is to know that when things become very unpredictable and volatile, they should look to reduce their time risk as much as possible if they do not have any specific alpha in the trade.
On the other hand, the volatility has created more opportunities for strategies that are posting passively, which helps clients to avoid the pain created by widening spreads in certain currencies. Gressel believes that clients that were willing to take on that risk have seen very good results in that timeframe. Other algo providers made similar reports to the BIS, who found that the most significant driver of increased execution algo (EA) volumes was a relative deterioration in other methods of execution. “In particular, risk transfer spreads widened significantly, especially for larger-sized orders, as market-makers were less inclined to take risk onto their balance sheets,” the report notes. “Meanwhile, passive EAs allowed users to manage their orders over time at prices inside the bid-ask spreads. Moreover, passive child orders filled more quickly as FX volumes picked up. Consequently, FX EAs – particularly passive pegged/tracker and time-sliced EAs – generally outperformed risk transfer during this period. They also had lower market impact compared with more aggressive limit-based/sweeping Eas.”
Conclusion
Looking ahead, Gressel says he would encourage clients to challenge their FX algo providers on the quality of the of their algos and the quality of the of the whole algo setup. He believes that as the FX markets may likely face further periods of volatility, it is important that clients understand how intelligent the algos are in assessing market environments. In addition, clients should also challenge providers to demonstrate that they have the appropriate controls and safeguards in place. “Finally, any client that is looking for an FX algo provider should seek out those who offer active support, with a team that is actively monitoring, working orders and can give feedback on those working orders,” Gressel concludes. “A reliable algo team should be good at assessing what has happened and be able to recommend if there are any better alternatives going forward to complete an execution.”
