The percentage of FX business done through algorithmic-based trading across all institutions rose from 27% to 33% between 2014 and 2015, according to the Greenwich Associates’ 2015 Global Foreign Exchange Services Study.
The rise was fuelled by investments in FX algo technology and the increased availability of electronic platforms and APIs.
In the context of a decline in non-electronic corporate FX activity in the UK and particularly the US last year, this indicates that the proportion of business done via algos has increased.
Guy Hopkins, head of sales for MahiFX’s MFX Vector product, says there has been increased interest from corporates – alongside a greater marketing effort around such solutions.
“This reflects the wider adoption of algos across all client types,” says Hopkins. “Corporates wish to minimize FX costs in the same way as other buy-side clients and are increasingly unwilling to pay a wide spread on an RFQ-type platform.
“Banks, and the salespeople within those institutions, are also looking for ways to differentiate themselves and a high-quality algo offering is an ideal way to do that. As a result, algos are being marketed much more aggressively.”
Aside from algos’ usefulness in day-to-day treasury operations, corporate customers have also identified they are potentially valuable when it comes to notable event-driven trades.
Experienced bank traders still have a big part to play in facilitating these transactions for clients, but algos are now being used as a way to work risk into the market without identifying the corporate’s intentions to other market participants.
The criteria used by corporates to select an algo strategy are influenced by specific best execution policies and objectives, explains Pete Eggleston, co-founder at BestX.
“It can be a difficult decision-making process due to the multiple factors involved and multitude of products available,” he says. “This is where consistent analytics can help more informed decisions to be made.
“Criteria may include execution style, benchmark objectives and cost, the last of which is particularly difficult to make a judgement on as net performance is more important than headline fees.”
Treasury desks tend not to be focused on adding alpha and are not compensated for taking risk, so strategies that track the market in some representative way – for example, TWAP or VWAP – are generally more popular than those specifically benchmarked against a point-in-time measure, such as arrival price. Corporates will often seek to be as passive as possible.
“The corporates we face typically have such large tickets that if they went straight to market, the market impact would be too great,” says Bryan Seegers, co-head of global eFX coverage at ADS Securities.
Algos have enabled trading to happen faster than ever before, but they are highly dependent on accurate pricing data.
Corvil chief business development officer David Murray notes that the most important aspect of pricing data is the age of a price, or the difference in time between when a price was first created and when the corporate receives it.
“Stale prices mean algorithms cannot execute optimally, so the age of a price is a critically important metric to account for,” he adds. “If the algo does not have sufficient data around price, or does not know how old a price is, it is not going to yield optimal execution.”
Banks have invested substantial amounts in post-trade transaction cost analysis (TCA) to demonstrate the execution quality of their algos, but there is an increasing realization that the provider’s TCA on its own is not sufficient.
According to MahiFX’s Hopkins, some form of independent analysis is crucial – not just to confirm the execution performance but also to put it into the context of the performance of other providers.
Paz at Aite Group observes that since a lack of pricing data will tend to cause algos to slow to a crawl, it makes no sense for a fast-moving “viper” algo to be let loose at a time of low pricing activity.
“It would be like flagging the market and saying that you want them to front-run your order and move the price against you,” he says.
There are also risk factors to consider. Algos are by definition complex automated trading strategies, and even seemingly simple strategies such as TWAP and VWAP require numerous layers of infrastructure, connectivity and expertise.
Hopkins suggests that users of algos cannot possibly understand all the intricacies of the different strategies they use.
“This is where rigorous, comparative assessment becomes absolutely paramount, to be able to prove demonstrably that the algos being deployed have been selected on the basis of performance,” he says.
Wolfgang Fabisch, CEO of b-next – a specialist provider of capital-markets compliance solutions – reckons a new approach to surveillance is required.
“In addition to protecting the market, industry and customers from fraudulent trader activity, we must also improve our understanding of algo activity,” he concludes. “We need to know how algos are programmed, by whom and whether they are being manipulated or attacked by external parties.
“This will extend the scope of surveillance into forensic work as well as IT security, and chief compliance officers will have to extend their remit similarly.”