Best Execution - Do Algos have a role to play ?
The use of execution algorithms in the currency markets has increased significantly over recent years(1) (2) (3) and it would appear that this trend is set to continue for the foreseeable future. Why is this ? In this article we explore the benefits of using algorithms, but we also review the potential pitfalls that users should be aware of if they are to incorporate the use of algos in their execution process.
Algo Types
Before we investigate the benefits and pitfalls of using algos, a brief overview of the range of algos now available for FX should be helpful. Although there are now well in excess of 100 different FX algos available on the main multidealer platforms, this universe can be simplified into a number of broad categories as illustrated in the diagram below. There are, however, different dimensions to how an algo should be characterised, for example:
Algo style (e.g. is the algo trying to achieve a specific benchmark or is it purely accessing liquidity opportunistically ?)
Liquidity source (e.g. is the algo only sourcing liquidity as principal or does the algo behave in an agency format via direct market access, or does it behave in a hybrid form of both ?)
Liquidity interaction (e.g. does the algo only aggress liquidity or does it also place bids/offers and interact in a passive way ?)
So, any given TWAP algo could be very different to another TWAP algo depending on the liquidity sources it has available to it and how it interacts with this liquidity.
Benefits
First we’ll review the potential benefits of including algorithms in your menu of execution options, which are summarised below.
Potential to reduce costs
Potential to reduce market impact, especially for larger tickets
Ability to access wider range of liquidity sources
Operational efficiency
Transparency and audit trail
Cost reduction
Using an algorithm can result in significant transaction cost savings, although there are a number of caveats here that we’ll cover in the following section on pitfalls. If an appropriate algo is selected for the specific execution objectives, benchmark and prevailing market conditions, then splitting larger orders, or less liquid orders, via an algo can result in cost savings. Such savings can result from a couple of sources: i) reduced spread costs as the individual child orders are smaller and, depending on the algo used, may in some cases actually result in not crossing the bid-offer spread at all, and ii) reduced market impact costs.
Market impact
Reducing market impact is becoming a higher priority for market participants as the increasing fragmentation of liquidity within the FX market, and associated volatility, is making execution of larger ticket sizes more difficult. Carving an order into smaller child orders, and distributing carefully via a smart order router, can help reduce impact, or market footprint, thereby resulting in improved overall execution. However, once again, caveats apply as poorly performing algos, and/or not very smart order routers, can result in significant signalling risk to the market, which may result in increased impact.
Liquidity sources
The increasing fragmentation of liquidity as new venues are established, and new participants enter the market, for example, non-bank liquidity providers, is creating a significant overhead for participants who would like to directly access as much liquidity as possible. The use of certain algos effectively provides this access in a cost effective fashion, as the algo provider is delivering and maintaining all the necessary venue connectivity.
Operational efficiency
Algos can be a useful tool to increase efficiencies by effectively outsourcing the management of certain orders, which may be beneficial to free up time for an execution desk to allow focus on particularly challenging trades in terms of size and/or illiquidity.
Transparency and audit trail
Another trend in the FX market, in part driven by the scandals associated with fixings and last look, is the desire for increased transparency. Using an algo should result in the delivery of an associated post-trade report which provides full details of exactly how each child order was filled, providing a full audit trail of execution prices and time stamps.
Pitfalls
So far so good ? Algos clearly have benefits and can help contribute to achieving best execution in some circumstances, but they are no panacea. There are potential pitfalls to be aware of when using algos, which have been summarised below.
Cost vs performance
Market footprint and signalling risk
Marketing spin and black boxes
Liquidity sources
Selection
Cost vs Performance
Algos are typically charged for in terms of a specified amount per million of notional, for example $50/M. Different products from different providers will have different costs, and the temptation is clearly to simply use products that appear to cost the least. However, a product that might appear relatively expensive in terms of headline cost, might on average deliver far superior execution performance when taking into account market impact etc. Thus, to make informed decisions when comparing products, it is important to look at net performance of a large enough sample of trades to make the results viable. This is a challenge when each provider offers only their own performance data. More crucially each provider generally self-selects their own metrics and presents them in a unique non-standard format such that it is exceedingly difficult to meaningfully compare any two provider's benchmarks and performance side-by-side without the dedication of a significant amount of time and energy.
Market footprint and signalling risk
As just alluded to, net performance is critical and a key component of this is how much market impact any one algo may or may not create. There are a myriad of ways that algos can interact with sources of liquidity, both in terms of the nature of the sources and the way that child orders are actually placed into the market. Poorly designed algos or sub-optimal selection and management of liquidity sources may create significant market impact, which may be compounded by allowing other market participants to further identify the algo behaviour through the signalling risk it has created. So, if a key execution objective is to minimise impact, it is not a given that an algo will automatically achieve this and indeed, depending on the market conditions, using old school risk transfer via the phone may at times be a superior execution method.
Marketing spin and black boxes
There is a bewildering array of algo products now available from many different providers and navigating the maze of marketing material is not straightforward. For example, on a platform such as Bloomberg, clients can access over 100 different FX algo products from multiple providers. Understanding exactly what any given product does, and how is does it, can be challenging as providers understandably are protective of their intellectual property. Conversely, there is growing pressure for users to have full disclosure on the details of how an algo actually works. The vast majority of market participants are uncomfortable using any product that might be perceived as a ‘black box’, and this is probably one of the reasons that the adoption of algos in FX is still largely concentrated in the more ‘simple’ product types.
Liquidity Sources
For algos that access multiple liquidity sources, this can create overhead in terms of monitoring which liquidity providers are delivering high quality liquidity and superior execution, through, for example, low reject rates and rigorous enforcement of participant behaviour. Algo providers manage these relationships although a client may have specific requirements in terms of where they want their algo orders to be worked and where they don’t.
Selection
Given the bewildering array of products now available, the process of how to select a specific algo for the trade in question is obviously complex. A number of factors need to be incorporated within the selection process, including if there is an execution benchmark to consider or other execution objectives. Clearly, it makes little sense selecting a TWAP algo over several hours if a trade has a specific benchmark of mid-market Arrival Price. Even within a certain genre of algo types, such as TWAP, how do you choose between providers ? With the increased focus on FX execution from asset owners, trustees, compliance and regulators, it is more important nowadays to be able to record and justify such selection decisions. A rigorous process, informed by independent analytics, provides the foundations for such a selection process by focusing on objective performance measurement.
Conclusions
Algos clearly have a potential role to play in any modern FX execution process, helping provide efficient, transparent execution in a cost effective manner. However, they are no ‘magic bullet’ and won’t necessarily always deliver ‘best’ execution. It is imperative that upon embarking with the use of algos that the potential pitfalls are understood and they are used judiciously and when appropriate to the trading objectives, benchmarks and liquidity conditions. It makes sense to include algos on any menu of execution options and products but it doesn’t make sense to always choose them for every meal. There may be times where the fish option is preferable to steak or vegetarian.
(1) Greenwich Associates, October 2014, reported that FX algo usage grew from 7% of clients in 2012 to 11% in 2013, with an expectation of this rising to 18% by end 2014
(2) Greenwich Associates, May 2016, reported that sophisticated investors executed 61% of their currency trades via automated computer programs in 2015, up from 33% in 2014
(3) GreySpark Partners, September 2015, estimated that algo trading will account for approx. one third of all currency trading in 2016