Shining a light on Covid-19 What has the buyside really learnt about FX algo use during the crisis?

As the data analytics, reports and whitepapers begin to drill down into the detail of FX algo performance during the height of the Covid-19 volatile market period, the true extent of increased buyside usage is now fully ...

Shining a  light on Covid-19  What has the buyside really learnt about FX algo use during the crisis?
Industry Views

Nicola Tavendale

According to analysis published by Refinitiv’s FXall, recent usage trends confirm algos have become an everyday tool for the buyside - and not just in times of low volatility. Algo trading volumes in March increased by 380% year-over-year on FXall, with the bulk of the increase attributable to asset management clients, while the continued heavy workload faced created by tough market conditions means the buyside is increasingly turning to smart tools to enhance their trade execution. 

A JP Morgan report provides further evidence for this increase, with more than 60% of trades for ticket sizes of more than $10 million having been executed in March using an FX algorithm, with a notable rise in use form hedge funds and real money accounts. In comparison to other periods of volatility, the weight of FX algo versus the use of traditional risk transfer was surprising, as during previous events or times of volatility clients had been unsure whether to use algos or not in those market conditions, says Carlos Gomez Gascon, Executive Director, Head of FX Algorithmic Execution at JP Morgan. 
“Our analysis with the clients and traders during this period of high volatility demonstrates that we continue to outperform the risk transfer price,” Gomez Gascon adds. “Another interesting finding is that because there’s more automation and the traders are doing more, what they need are just the highlights. As a result, we’ve been experimenting with new workflows and have also developed a new alert called VPIN (volume synchronized probability of informed trading) which is based on the findings of an academic paper which found that a VPIN alert could have predicted the SNB flash crash hours before it happened.” 

The aim of the alert is to warn JP Morgan’s algo users if there is some probability that a market event is likely to happen or that there may be extra volatility, Gomez Gascon explains. This helps reduce the need for clients to constantly monitor the algo in flight and can instead rely on there being an alarm in place which will warn them if they need to turn their attention back to the algo.

During the crisis, the buyside is also very likely to have faced its own set of constraints without their usual operational set up. Gomez Gascon adds: “In those circumstances an algo may prove to be more efficient for their workflow, freeing them up to focus on other areas where they can add value.” 

Algo Covid

Ralf Donner

“We also expect to see further migration towards some of the newer products that clients saw were particularly helpful, such as the basket algos.”

Ralf Donner, Global Head of Client FX Algo Execution at Goldman Sachs, says the value proposition for algos was even getting reduced in last year’s quieter markets. Although clients were getting more comfortable using algos in that low volatility period, he says that on the other hand the competition from risk transfer was extremely tight to the extent that it was difficult to justify using algos in some cases. 

Resilient and reliable

“What the crisis did was it brought that volatility back and interestingly, instead of migrating away from algos, we saw a huge uptick,” Donner adds. “Of course, the conditions were measurably worse than pre-crisis and yet it was still competitive, because other alternatives to algo execution were also significantly more expensive. For many clients, the fact that these algos did perform comparatively even under more difficult market conditions has really helped to boost confidence in their use.” Yet now that the alternatives to execution have started to get cheaper, Donner notes that they are still seeing higher algo volumes than they had last year, which he says suggests that the legacy of the crisis has been to provide clients with confidence in the product.

FX algos have also proven beneficial to clients in these difficult trading conditions, as they are used to dealing with fragmented liquidity conditions. This means that even if liquidity dries up in one place, the algo can reroute instantly much faster than a human trader could, Donner says. In terms of which strategies proved to be the most effective, he notes that his team saw an uptick in volume for nearly every category of algo except the most aggressive kind, the sweep-to-fill type of algo. “Then there were two algos which really stood out, the first being our flagship dynamic hybrid algo. This was largely due to a fortuitous investment that we had made in the previous 12 months in adaptive dynamics, or self-learning logic in other words, for the algos,” Donner says. That enhancement meant that there were fewer hardcoded, calibrated parameters, which gave more ability to the algo to update on the fly, which greatly enhanced its performance. The second strong performance was in Goldman Sachs’ basket algo type products, which are better for trading highly correlated markets. “They had a number of successes. It’s a small but loyal following that we’ve got for the for the basket algo  and we certainly saw larger ticket sizes on basket algos and a strikingly good performance there,” Donner adds.

Algo Covid

Vittorio Nuti

“Now clients trust the product more than ever before after they’ve seen how well the FX algos perform in both periods of low volatility and high volatility,”

FX algos appear to have proven to be as resilient in volatile markets as providers have long assured clients they would be, but significantly have shown that the checks and balances are also extremely robust. It is vitally important to have the ability to slow down or interrupt execution depending on the market conditions, explains Vittorio Nuti, Global Head FX Algos at Deutsche Bank. “I believe our balance overall was the correct one and we didn’t have any major issues,” he says. “Now clients trust the product more than ever before after they’ve seen how well the FX algos perform in both periods of low volatility and high volatility, which helps a huge amount in building confidence.”

Looking for liquidity

According to Nuti, the feedback from clients on the effectiveness of the analytics offered by Deutsche Bank, such as the Market Colour app, has also been extremely positive in terms of the ability it provides clients to see and understand what is actually happening in the market and how the execution is going in real time. Yet while he says that passive strategies do tend to give clients the better results, as they take more market risk, as the volatility has subsided, clients are also still comfortable with the more aggressive strategies to clear risk. “Those are the elements which are important to clients,” he adds. “Yet a lot of clients during the periods of high volatility had varied experiences in terms of the liquidity provisions that each algo provider has, with some finding it very difficult when using other providers when the liquidity disappeared quite quickly.” That wasn’t the case at Deutsche Bank, where Nuti says the team worked hard to manage the liquidity and to make it a fair provision for all parties involved, which means ensuring there is liquidity when things get busy as well. 

Nuti explains: “We offer multilateral liquidity, like the ECNs, and then we have bilateral liquidity which includes both internal and external flows. It is only the top quality LPs that provide liquidity to our algo business and we work closely with them to make sure that they provide high quality liquidity but also that our algorithms are going to be trading appropriately on those feeds as well. It is very much a two way street. That sort of partnership really helps when you have moments of stress because it is viewed as a more longer term relationship rather than just a variety of ECNs or anonymous feeds, where pricing could just disappear all of a sudden, which happened to some of our competitors.”

In addition, the overall levels of technical support available faced a particular challenge during the height of the crisis, with both buyside clients and the teams of FX algo providers having to adapt quickly to working from home. The primary focus has to be sustaining that high level of service provision to clients, says Loïc Bourgeois Ducournau, e-FX Liquidity Sales at Societe Generale. “We had to ensure consistency even though of course we were also very, very busy,” he adds. “The challenge was to pitch new clients who perhaps did not want to risk using any new tools or algos when they were working from home, but those clients who were already comfortable using algos before Covid were very happy with how they performed and the support provided, even while they were working from home.”

Algo Covid

Carlos Gomez Gascon

“The use of algos is going to continue to increase, because there is going to continue to be a focus on cost reduction and scalability,”

Continued use

The increased volatility also meant that volumes exploded during the volatile Covid period and according to Bourgeois Ducournau, the levels of algo adoption are continuing to increase even now. “That is in keeping with the market trend, with some clients preferring to pay fixed fees and have full transparency instead of to making that trading information visible to the Street,” he adds. “It’s really, really efficient. It also provides clients with the ability to have much more access to liquidity and not only bank liquidity. Clients are looking to passive algos in order to spread their execution.” 

However, Bourgeois Ducournau adds that during the volatile period, they noticed that some clients preferred to do risk transfer price and to get out of the risk very quickly, instead of risking a very sharp and strong market move. Then on the other hand, they also saw other clients who were very happy with the algo execution and tended to prefer aggressive execution using the Falcon algo, which executes very quickly to get out of the risk as quick as possible. “It works very well in liquid pairs and we had some of our more experienced clients using it, with strong performance results,” he says.

Although the markets are not in that highly volatile period anymore, the new trading environment we’re now in is set to continue in the foreseeable future. Gomez Gascon says that regardless of this new Covid world, the point of using algos is for scalability of execution and is going to be increasingly recognised. “Buyside clients can start focusing on the real value which algos can bring,” he adds. “Algos are a useful toolkit for them to focus on adding the most value on the strategy. While they work from home, algos mean that they don’t need to be doing several things at the same time, they can just let the algos run and then set up the alerts that they think are interesting or in case they need to  intervene and so  can focus more on strategy. The whole process is seamless.”

Algo Covid

Loïc Bourgeois Ducournau

“Those clients who were already comfortable using algos before Covid were very happy with how they performed and the support provided, even while they were working from home.”

Turning to sophisticated strategies

FX algos also offer an important benefit at a time where the cost of doing business is under the spotlight as the automation piece also offers considerable savings. According to Gomez Gascon, it’s not just about the quality of the algo but about all the things that surround it that actually allow the buyside to save money. “That’s why the use of algos is going to continue to increase, because there is going to continue to be a focus on cost reduction and scalability,” he adds. Even during the highly volatile period, there were also no unexpected risks uncovered and as Central Banks now start to evaluate the market response to those conditions, there are on the whole no unpleasant surprises to report. Donner explains that while the markets were volatile, they were typically well behaved and did not really run the risk of major outages or something going very badly wrong.  

Even so, having algo performance tools to monitor algo performance live proved essential, Donner says. Especially when market conditions are unusual, he says that clients were able to rely on having a live analytics tool to monitor where the liquidity was. So if the primary and secondary markets have evaporated, then the performance of the algo is going to go down to internalisation opportunities and any mid book matching that’s possible, Donner notes. Analytical tools that were less useful in this period were anything with a long look back window, perhaps calibrated to other market conditions or historical datasets of executions, those were less likely to be relevant.  

It is also unlikely that providers will continue to hear that buyside clients only use FX algos in ‘no vol’ environments, Donner adds. “Many clients have come to the conclusion that there was a good performance also in the higher vol period and we expect they will be emboldened to continue using the product because they’ve seen that it works under quite difficult circumstances,” he says. “We also expect to see further migration towards some of the newer products that clients saw were particularly helpful, such as the basket algos. The buyside has now come to the conclusion not just to use FX algos, but hopefully that they will start using some of the more sophisticated algos that are smarter about executing things together.”