Closing the information gaps – Helping the buyside use FX algos more effectively

Last year’s BIS report on FX execution algorithms (EA) identified closing information gaps – both in terms of expertise and data – as being a central area of focus needed to ensure greater transparency and a level playing field for market participants. As regulatory bodies, such as ESMA, and the GFXC, in the context of the Global Code review, also explore ways to improve the information available to users, what can providers be doing to help the improve their clients’ ability to better assess the quality of their algo executions? Nicola Tavendale investigates.

Nicola Tavendale
Nicola Tavendale

As the three year review of the FX Global Code progresses, we will likely see that the industry as a whole believes that buyside participants will benefit from increased transparency of their algo executions, says Nickolas Congdon, Head of eTrading Services at Commerzbank. One way to achieve this aim which is currently being explored by the GFXC is the possible introduction of a standardized template for transaction cost analysis (TCA) reporting and common due diligence questionnaire templates which aim to provide uniform bank disclosures. “Through this, the buyside will be more equipped to understand how their execution was managed and easily compare strategies across multiple bank providers with a clearer view as to what and how their service operates,” Congdon adds. “This will allow the buyside to be more informed as to the key execution decisions each strategy employs and how the smart order router (SOR) decision logic works for venue selection, order placement and factors for execution.”

“One key indicator we look at across venues is the rate at which liquidity is throttled (replenished) because when liquidity is taken from one venue it is likely throttled or withdrawn completely from other venues” – Nickolas Congdon

However, Congdon warns that access to market data and objective analysis can be costly and appropriate performance analytics should cover the entire trade life cycle. “Where possible, the buyside should try to compare their overall execution against objective reference data and not necessarily rely on the benchmarks provided from a single provider,” he adds. According to Congdon, almost all third-party TCA providers already allow the buyside to compare the performance of the execution algorithms offered across a wider aggregated execution data set than just their own, which enables them to identify top performing strategies without unnecessarily paying for performance as they learn.

It is also becoming increasingly essential that traders should understand the objective of each strategy and how it makes routing decisions, says Zeke Vince, Global Head of eFX Sales at Bank of America (BofA). He explains that at a parent order level, performance against standard benchmarks (Arrival, Risk Transfer, TWAP, etc), venue breakdowns, impact and reversion metrics are all invaluable. “It is also useful to see an audit trail of placements and executions to help drill into strategy decision making to better understand its mechanics,” Vince adds. “Firms looking to assess algo execution should review parent-level performance against various benchmarks, spread capture and distribution of fills to help gauge an algo provider’s access to unique liquidity and the routing logic, which underpins liquidity access. Increasingly, firms are using this data to help direct more of their algo execution in a ‘wheel’ type of strategy.”

Assessing the data

“The main risk you take when using an EA is duration risk (thus market risk) versus trading against a principal channel,” – Zeke Vince

In addition, it is now a common standard for banks to provide high quality execution reports, as well as using the services of independent providers, to help clients understand the quality of execution, says Ian Daniels, Head of eFX Distribution EMEA at Nomura. Although TCA reports are not indicators of future performance, Daniels explains that they can still provide good insight into the value added by an execution algorithm, such as by providing an average slippage for comparative orders and its distribution, for example. “It is also certainly worth seeking independent providers of execution reports to challenge the existing methodology of your providers,” he adds. “History tells us that many negative market surprises came from ‘correlations’ – all trading the same direction, everyone using the same risk metrics, everyone believing the same narrative.”

One preferable way of utilising TCA reports is to consistently compare them with pre-trade reports, notes Daniels. “Any systematic inconsistency you find, especially when predicted pre-trade outcome is generally better than post trade execution, may suggest the need for an improvement in your execution algo portfolio,” he warns. Real time reports, on the other hand, can be the most challenging to consistently use, according to Daniels. He explains: “Even though it may be tempting to improve an algo in flight, it can be difficult to assess whether any such intervention had a positive impact while avoiding any cognitive bias.”

A further area that is increasingly highlighted by algo users is the need for better information regarding the liquidity sources being accessed by the algos. Mary Leung, Global Head of Client Algos at State Street, explains that this due to the highly fragmented nature of the FX market today, with the number of trading venues, ECNs and liquidity providers having significantly proliferated over time. Leung adds that this fragmentation can give rise to a “liquidity mirage,” which occurs when liquidity providers are market making across different venues simultaneously to gain market share. “An algo provider may assert it has ample liquidity, with connections to 10 or more venues. However, behind the scenes the liquidity is actually coming from the same small group of providers,” she says. Adding to the complexity, each venue plays by its own matching rules, according to Leung. Some offer a central limit order book (CLOB), while others allow curated liquidity pools, then others have either a ‘firm’ or ‘last look’ liquidity protocol, or both, while some choose to restrict providers and participants.

Looking at liquidity indicators

“Simply put, the higher the internalisation ratio, the lower the expected market impact, because information leakage and signaling risk to the market should be reduced.” – Mary Leung

“Because of the anonymity of these ECNs and CLOBs, algo providers can only run venue analysis on the ECN level rather than on a more granular liquidity provider level. The result is that different algo providers may be connected to the same ECN, but may not necessarily have the same liquidity access or quality,” Leung says. At State Street, Leung explains that these issues are addressed by using disclosed direct connections as the main choice of liquidity curation, allowing the bank to interact with its liquidity providers bilaterally and transparently. “This allows us to have full control over which end providers we curate in our liquidity pools at any time and we can gain a clearer picture of the quality of the liquidity through spread information, market impact, response time and fill ratio of each provider,” she adds. “This transparency supports higher fill ratios overall, compared to a typical ECN order matching process.”
Even so, Leung notes that despite their challenges, ECNs continue to play an important role in sourcing liquidity. State Street connects to a select group of proven, high-quality trading venues for passive orders and other liquidity needs, complementing its direct provider approach, she adds. “We are highly transparent about the sources of our liquidity and can share this information with clients upon request, while we also welcome the opportunity to engage with clients who want to be more involved in the decision-making process and have a greater role in curating their own liquidity pools,” she adds.

Ultimately it is the algo providers unique access to, and management of, liquidity that forms one of their key value propositions, argues Congdon. He believes this should be a healthy mixture of internalisation, firm, disclosed, anonymous, passive and aggressive based liquidity from a myriad of sources. “Users should not only be aware of where their provider is executing, but also what liquidity sources the EA has access to,” Congdon adds. “Just because a fill is from firm liquidity or internalised, doesn’t mean it was the best fill for the client.” He explains that historically price was the key driver when accessing liquidity across numerous pools, but now many more factors such as order book depth, priority, fill probability for non-firm venues and the source of liquidity should all be weighted. Each clip should also be compared to a neutral independent mid-price at the time of the fill, Congdon says. “This is another area that standardised TCA benchmarking and due diligence questionnaires on how providers manage liquidity would add value for the buyside,” he adds.

Improving insights

“Once you have built a level of trust in the algos themselves, then it makes sense to leave them to run and not to try to manually improve the performance,”..- Ian Daniels

Additionally, Commerzbank’s order service is data driven backed by decades of internal investment into proprietary technology. This translates over to the buyside, Congdon explains, which allows clients to be more informed as to how Commerzbank monitors and manages liquidity conditions in a highly fragmented marketplace. “One key indicator we look at across venues is the rate at which liquidity is throttled (replenished) because when liquidity is taken from one venue it is likely throttled or withdrawn completely from other venues,” he adds.

Tan Phull, Head of FX Algos at BofA, adds that when evaluating internalisation of flow, the buyside needs to consider where it is versus other client flow or principal liquidity. He explains that this distinction ought to be clear and customers should be given a choice to opt in or out of these forms of liquidity, as well as being provided with the information needed to understand the benefits and limits of both options.

“Increasing fragmentation has been one of the primary problems that algo execution seeks to address. A more fragmented market makes it imperative for broker strategies to route optimally, in order to reduce impact and source appropriate liquidity,” Phull says, adding that merely knowing the venue from which the liquidity is sourced is no longer enough. Instead, he believes that buyside firms can better inform themselves by consuming data that tells them where within the spread they executed and if they provided or took liquidity. “At a parent order level, a summary by venue is important to understand the makeup of the execution, as well as post trade reversion metrics at a venue level. This information is available to our clients via our Real Time TCA portal,” Phull adds.

Understanding liquidity in the FX markets has always been a challenge, Daniels adds. Different platforms and providers allow the buyside to source market colour and help them to understand the flow dynamics, he explains, adding that it is worth also following the imbalance between different sectors and trying to understand the flow between those sectors. “Indicators incorporating the depth of the market also provide invaluable information insight,” says Daniels. In addition, he believes that the execution reports for past algo executions can also provide clients with a greater understanding of liquidity trends in the FX markets. “In particular, if a firm regularly executes similar trades, spotting the trends within the execution report can indicate improving or deteriorating liquidity conditions,” he says. From the EA perspective, Daniels notes that it is also important to be clear on the objective (what metric to use when assessing the EA itself) and then adapt the choice of the EA. “Good algos adapt to current liquidity conditions and keep the market impact under control,” he adds. “Once you have built a level of trust in the algos themselves, then it makes sense to leave them to run and not to try to manually improve the performance, as the algo could have better access to current market conditions than the user.”

Defining key concepts

A further important factor that clients should consider when evaluating an algo provider is the internalisation ratio, says Leung. She explains: “Simply put, the higher the internalisation ratio, the lower the expected market impact, because information leakage and signaling risk to the market should be reduced. Recently, our clients have expressed concerns that comparing these rates across different algo providers is becoming more difficult, due to various industry definitions of ‘internalisation’.” The source of confusion lies in the fact that while most algo providers only count trading against their own internal desk as internalisation, others may extend the ratio to include fills sent to external bilateral market makers, notes Leung.

Also, she warns that not all types of internalisation are equal in terms of their market impact and information leakage. “For example, a principal trade to a provider’s own eFX market making desk does not always mean it’s ‘internalised’ because it depends on how that desk ultimately manages the position. The trade can also lead to indirect information leakage,” she adds. Leung says that although skewing prices to attract flow is a popular practice amongst some providers, when it is done via skewed streaming prices, it may give out more information to the broader market than intended. To combat this, State Street has recently rolled out a concept that changes the mechanics of skewed streaming while trying to achieve the intended results. “Thus far we have seen great results in terms of improving algo execution, reducing market impact and decreasing information leakage,” Leung says.

Leung warns that clients should also have a good understanding of the definition of ‘internalised’ that is used by a provider and therefore the kind of internalised liquidity they are truly matching with. “At State Street, we provide our clients with information on different internalisation options, our sources of liquidity, internalisation availability and allocation logic among competing orders,” she adds. “We also offer complete transparency around how internal liquidity types are matched through our real-time and post-trade transaction cost analysis (TCA). All of this helps our clients to make more informed decisions and improve their execution.”

It is now a common standard for banks to provide high quality execution reports

A shift in dynamics

If conducted appropriately however, internalisation provides buyside traders with access to unique liquidity with higher fill rates and in larger (block) size, says Vince. He explains: “BofA has a global customer business across corporates, real money, hedge funds and payments, so we are fortunate to be in a strong starting position. The net benefit to a client should be lower footprint costs and thus reduced market impact. A liquidity provider should be able to evidence this interaction through fill transparency and clear governance routines around the interactions within their internalisation engine.” In addition, buyside firms should be aware that, by definition, algos have a non-zero duration, Vince explains. “The main risk you take when using an EA is duration risk (thus market risk) versus trading against a principal channel,” he adds. “Our execution data supports the use of EA’s within a framework that outperforms the risk price in a number of scenarios, of course with some caveats.”

Congdon adds that traditionally, execution between the buyside and bank providers was conducted predominantly on an OTC principal basis, with liquidity providers taking the credit, settlement risk and market risk. Now however, through the use of algorithms, he explains that bank providers are able to offer clients the operational efficiency of centralised settlement and credit risk while remaining a legal principal, yet negating the market risk as a riskless principal. “The buyside must be comfortable with the inherent risks of their execution and intimately understand the trade-off between execution certainty and market risk,” Congdon says. “Different types of EAs respond differently to market conditions and are situated accordingly within the efficient frontier. With standard disclosures highlighting the key risks and participant liabilities, the buyside will be more comfortable allowing them to determine whether they are being adequately compensated for the additional market risk they are holding.”

He adds that it is also important to remember that algorithms are a tool to be used in the right circumstances, allowing the buyside access to deeper liquidity and providing operational efficiency. “The buyside hold the risk and as providers we need to ensure all clients understand the role in which we act and that this is clearly communicated in our disclosures and due diligence templates,” Congdon adds. “The improvement to Principle 18 of the FX Global Code will help the buyside to digest the service each provider offers. Both the TCA and FX Algo Due Diligence template would help navigate what can be a cumbersome and complex process of reviewing disclosure documents.”

A new standard?

Execution algorithms also allow a buyside trader to express a market view with speed, control and discretion over liquidity selection, Phull notes. In addition, he explains that BofA helps clients analyse their TCA data in order to customise the EA’s used to better suit their trading needs. He adds that these tools were created with a view to supplementing the provision of risk capital via principal and voice channels that BofA provide to clients across market conditions. “The industry would also benefit from a standardisation of disclosures required around EAs and their associated liquidity access,” Phull says. “Within that framework, liquidity providers should be able to innovate to improve their products. This makes for a healthier and more transparent market, which encourages continued innovation.”

Daniels adds that there already seems to have been a natural market unification and standardisation when evaluating EAs. He explains that EA providers across the Street use similar metrics, which gives buyside firms a wide range of information. Daniels adds: “One aspect of the EA where a unified approach may help is in terms of stress tests and behaviour of EAs in adverse market conditions, i.e., what if in the middle of the running TWAP there is a news event? What should we expect? What is the new normal now? Having an ex ante unified framework, where some expected outcome can be assessed, could mitigate many unpleasant conversations and give buyside firms a better understanding of the risks.”

Conclusion

Generally speaking, market participants still tend to face an uphill battle when trying to make informed choices and conduct proper evaluations of algo strategy logic, warns Leung. She argues that this is because many providers view this logic as proprietary and prefer to closely guard their intellectual property. “In addition, other topics such as liquidity provisioning, child order routing, kill switches, information segregation from the principal sales/trading desk are all crucial to the algo product offering, but can lead to conflicts of interest if not properly managed,” Leung concludes. “Every client should be asking for robust disclosures from all their algo providers to ensure algo logic, and any conflicts of interest, are thoroughly transparent.”

Related Posts

Scroll to Top