The reality for traders on an FX desk is that when given an order to trade algorithmically, they are likely to only have a relatively small amount of time in which to make a decision about which of the myriad number of algos now available to use as well as from which of the rapidly growing number of FX algo providers in the market. As a result, anything that allows them to reduce the time between receiving the order and making the right decision is definitely going to have a positive impact on the trading outcome itself, says Mark Goodman, Head of Electronic Execution FX, Rates and Credit at UBS. “When we think about simplicity, we are taking an empathetic view. What conditions would make it as simple as possible for that client to make the right decision and achieve the best outcome as quickly as possible?” he explains.
According to Goodman, in addition to the timeliness of placing the order and the ability to make the right decision about which algo to select, a further key challenge is that the FX algo market is becoming an increasingly crowded space. “Whereas two years ago you might have had a selection of three maybe four providers, there’s now eight to 10 offerings in the market,” he says. “And the number of different offerings available from each of those brokers is also expanding, which in turn makes the selection process harder. Even experience traders can be dazzled by too much choice.” As a result, the growth in the range of algos and providers has unintentionally only made it harder for the client to distinguish between the different offerings and different parameters available.
Striking a balance
A further advantage in offering clients greater simplicity when it comes to their algo selection is that this also makes it easier for them to explain their decision to senior management and investors. Ian Daniels, Executive Director, Head of eFX Distribution EMEA at Nomura explains that clients such as a real money firm or an asset manager, now need to be able to explain to their investors, PM’s or head of trading why they chose to execute in a particular way. “This needs to be simple to explain,” says Daniels. “Clients now have a plethora of algos to use. At last sight on certain third party platforms there were around 20 banks actively offering execution algos. So as a provider, if you make it too complex then a big part of the market will not choose you and will opt for someone with an offering that is easier to understand. That’s why at Nomura we aim to make this choice as clear and simple as possible.”
Nomura’s own FX algo suite currently offers a selection of five main algos. This ranges from a simple TWAP, which remains a relatively popular choice, in addition to Satori which is effectively a VWAP algo. To keep the selection as simple as possible, algos are named in such a way that they ‘do what they say on the tin’, Daniels adds. For example, the bank offers an FX algo called Ninja, which is more stealthy and passive, alongside one called Samurai, which is a bit more aggressive but still intelligent. “It’s not just a rudimentary algorithm, it is intelligent in the way it interacts with the market in real time while also able to look back at historical data,” says Daniels. “Then we have a more aggressive algo which is called Shogun, where the client is able to set the price limit where they want to execute and then the algo will sweep the available liquidity. Even so, it also has the intellect behind it to try and get some of the fills done passively below the limit price and then, once the limit price is breached, it will go and sweep the market. This will naturally have a bigger market impact than some of the other algos on offer.”
Even with a clearly defined set of algos which are named to match their characteristics, clients who are new to FX algos will still need a convenient way to choose them based on their objectives and compliance requirements for the execution. According to Paul Goldberg, Algo Execution Desk Head at Citi, simplicity in the form of wizards and other pre-trade tools should allow clients to easily map their execution objectives to an appropriate algo strategy, while limiting the number of options available on an algo will further help reduce the barrier to entry for many who may be migrating from RFQ, streaming prices and simple orders. “Adding a lot of options is an easy way to satisfy the custom needs or requirements of many clients,” he adds. “However, this makes it more difficult to consider different types of algo execution as part of a coherent set and thus makes it harder to infer statistical significance in performance when comparing across strategies.” Goldberg warns that now clients have a larger number of FX algo providers to choose from, they are increasingly dividing their executions among competing providers to compare performance. Yet by using simpler algos, providers will be able to apply optimal settings to run the algo and this can be used to better compare performance across providers, he explains.
Leave complexity to the engineers
“Not all clients will be able to optimally utilise all of the options available on some algos, due to the nature of their execution requirements. Very often, clients have a view about how they would like the strategy to execute. It may be the case that having algos with very simple controls may not allow them to map their idea of how the algo should work to something provided,” argues Goldberg. “We want to give clients the freedom to determine the style of execution, while ensuring that they can benefit from the core functionality that comes as a result of our quantitative research and from executing against the Citi franchise.” There is also a fine balance between making the ground too steep or going too far and making things oversimplified, while also keeping the controls as easy to understand as possible, adds Goodman. “The more you simplify the less control potentially the client has over the expected behaviour,” he explains. “That’s the balance, that’s the tightrope you’re trying to walk.”
Instead, algo providers need to consider how to give clients enough choice that they can properly reflect their own objectives into the algo behaviour, but not so much choice that they’re almost getting to a granular level where they’re almost driving the algo behaviour itself, Goodman warns. He adds: “Our view is always to ask: what is it that drives the performance for that client? What do we need to understand from the client in terms of their objectives in order to drive the right behaviour in the algo? If you get that balance right, you can give the client the right amount of control without getting overly complex. But if you get it wrong, you either get too complex which impacts the decision making process, or you get so simplistic that you’re almost giving just a one-size-fits-all solution to the market. If there’s one thing we know from dealing with clients is that every client is different, every situation is different and you need to reflect that in the offering.”
But is there now an argument that clients should be able to understand the inner workings of FX algos and how they have been engineered? According to Goodman, it is important to first distinguish between the simplicity of selecting an algo and its parameters and the actual engineering of the algo itself. In his view, clients are not asking for greater simplicity when it comes to how an algo has been built, but rather that they are focused on achieving the best possible outcome. “So many clients want us to use techniques which can be relatively complex,” Goodman explains. “We’re using a lot of machine learning in terms of how we build quant parameters, how we build signals into the algorithms and that drives good performance. But clients also want simplicity in terms of using the algo itself and in how they select which parameters to use.”
Appetite for transparency
Over complication can also confuse clients, warns Daniels, adding that the feedback from most of the bank’s clients is to “keep it simple”. To this end, Nomura offers its clients a ‘cheat sheet’ which provides a summary of what each algo does, alongside a short guide with clear and simple definitions. “Some clients also want additional information such as venue selection, our liquidity policing and what happens if a venue rejects an order,” Daniels says. “This can actually be key to certain clients having confidence in your algos. We’re in quite a fortunate position at Nomura as we are able to get access to various liquidity sources that are relatively unique. So it is important to us that we have policed that liquidity and keep the venues that now utilise our algos very clean, which in turn means our FX algos continue to perform extremely well.”
On the other hand, as long as the algo performs well or as long as the TCA looks pretty good or on the money then many other clients will not be too worried about understanding the inner workings of an algo, explains Daniels. “It really depends on their individual requirements. If it’s important to them, then it’s important to us,” he adds. “Even so, it is still important to have a good set of examples to go back over with each client in order to help them see how the algos performed. If they are just doing one or two algos, that is not a great barometer of performance, so they really need a good sample set.” Sometimes that can be one of the main reasons that clients actually struggle to see the benefit of adopting algos in the first place, he says “Some of the third party providers who have done this work on monitoring algo performance are actually helping out in this arena, because it allows clients that are having a practice run or have only used a few algos to easily see if it was worth doing in the first place,” adds Daniels.
According to Goldberg, clients also seek a fundamental level of understanding of how algos execute, particularly in light of certain characteristics: does the algo internalise, is it passive or aggressive, what constraints are there on my execution and when will the execution terminate? Citi, for example, provides post-trade TCA reports which provide child order level granularity around execution behaviour and help to confirm that the above execution characteristics are ensured. “Transparency is a key principle of our FX business and we are happy to provide further details about the inner workings of our algos on request,” Goldberg adds. “Pre-trade analytics are also now a feature of more sophisticated algo platforms. These types of analytical tool allow algo users to gain an insight into the estimated performance of an algo provider’s strategies based on historical data, often including duration estimates and slippage statistics against benchmarks such as inception mid and risk transfer price.”
Sophisticated but easy to understand
Naturally clients will also still require a range of toolsets which enable greater on the progress of their executions and insights into how their FX algos are performing. Goldberg agrees, adding that many algo providers are now incorporating real-time analytics into their algo products which in turn enables clients to track in real-time the progress of their algo at a child order level including order type, venue and price. In addition, providers are now actively improving the ways clients can access the algos and ensure they are able to integrate them with ease into their existing infrastructures and workflows. Citi, for example, has extended the availability of its full suite of algos and all available options to all major multi-dealer platforms and API clients. Goldberg explains: “This has made algos almost universally accessible, particularly for clients who cannot use a single dealer platform.”
One area, however, where clients are now quite sophisticated is understanding the inner workings of the algos because they’re trying to make sure they pick the right algo for their objective, adds Goodman. Yet he also warns that simplicity shouldn’t necessarily translate to very simple logic where you risk sacrificing performance. Goodman explains: “Take a very simple example of a TWAP. A TWAP could be built as a very simple algo so you’re just time-slicing the order but you can get more complex by actually giving the algo some discretion to take better advantage of the signals in the market to get slightly better performance. So clients want that sophistication within the algo, but that doesn’t mean that you call that TWAP something else. You call it TWAP and you put the right parameters on it and that makes the selection simple - even if the underlying engineering is complex.”
Goodman adds that at UBS, they have always been impressed with the level of sophistication clients want to get into in terms of detail. As usage has increased, so too has the demand for greater transparency and knowledge. “Simplification, if you get it at the right level, is already a selling point,” Goodman argues. “But if you make it too complicated, or if it’s not intuitive for the client, then you’re probably already losing market share.” Daniels agrees, adding that providers who can demonstrate consistency of performance and share similar client experiences with newer clients are also helping to breed confidence in algo trading. “If you look at Asia, for example, those markets are a little bit more in their infancy,” he says. “As an Asian bank, it’s important for us to be in constant communication with clients prior to the actual execution, in-flight and also post the event. We’ve also been able to call upon our experiences in the Americas and EMEA to help set some of the client expectations on algo usage.”
Recognising the advantage
Yet just how likely is simplification likely to become a key area of differentiation amongst FX algo providers in the future? There is always a market impact vs speed discussion to be had, particularly in a very fragmented markets like FX. According to Daniel’s that, coupled with regulatory drivers, has been one part of the reason for FX algo growth in recent years. “Market impact, and hence performance, is key,” he adds. “From client demand, we’ve seen that simplicity is breeding confidence which makes it easier for us in turn to help change trading behaviours. That has been a key factor behind our ability to grow and increase the penetration of our algo suite.”
Finally, a key factor which may help to promote further knowledge sharing between providers and clients is likely to be the influence of the FX Global Code. Citi, for example, were one of the first signatories of the Code, a key element of which is to promote effective communication that supports a robust, fair, open, liquid, and appropriately transparent FX market. In addition, conferences, workshops, whitepapers, disclosures, client visits are all key to efficient knowledge sharing and better clarity on algos, says Goldberg. “At Citi, we hold regular events which help bring our algo clients together so that they can benefit from not just our knowledge, but the collective experiences of their peers,” he notes.
Clients are also becoming more sophisticated and are interested in a more data driven approach to execution and measurement of performance of algos, Goldberg claims. “The more specialised customisations are made available for the algos, the sparser the data set becomes for each algo customisation. This in turn decreases the ability of the providers to optimise the full algo offering,” he says.
As the historical dataset on algo executions continues to grow, these can also be used within pre-trade TCA in order to steer clients towards better outcomes, based on historical executions in similar conditions. Increasingly, clients are asking for simplicity in algo choices as they are converging on a small number of ways to execute.
These can be either benchmark driven (VWAP/TWAP/implementation shortfall), style driven (passive/aggressive), or venue driven (external/internal/firm/last-look). As a result, algo providers who have gained the greatest expertise in optimising the execution of a small suite of easy to understand algos will be at an advantage, Goldberg predicts.