Execution Algorithms - Slice, Dice and Rejoice?

Abstract
FX markets have witnessed a rapid technological overhaul for the past few years. Digital
transformation of the FX value chain over the years has led to sophisticated pre-trade
analytics, electronic execution mechanisms, innovative risk management and highly
automated post-trade workflows. Technology and analytics are no longer seen as a side
function but as a driving force for business and trading transformation. Algorithmic trading
is an exciting emerging field that has evolved rapidly in the last few years, especially in FX
spot trading. There has been a strong trend towards greater fragmentation in the FX
markets, and execution algorithms (EAs) are emerging as tools to help users by aggregating
liquidity and facilitating access to various liquidity pools, which would be impossible
manually. EAs can help users reduce market impact and cut-down transaction costs while
improving execution consistency and fulfilling best execution requirements. However, EAs
are no silver bullet, and their usage gives rise to unique risks and challenges that warrant
close monitoring.

What is Algorithmic Trading?
At the most basic level, algorithmic trading entails the usage of a computer program
following a predefined set of instructions to place a trade. However, “algorithm” is an overloaded
word whose meaning depends on context. Most folks tend to think of algorithms as
top-of-the-stack investment strategy making investment decisions like order timing, how to
enter or exit position etc. However, there are two additional important layers, i.e., the
execution algorithm (EA) and the smart order router (SOR).

For example, let’s say a hedge fund strategy runs inside an algorithm and decides to buy 500
million EUR/USD to open a position. That would represent a parent order from the
investment strategy. Since the order is too large and placing it directly on the market may
create an adverse market impact, it’s handed over to an EA. The EA would typically work on
the order for a few minutes and slice this large “parent” order to generate multiple smaller
“child” orders. These child orders would then typically be passed over to the third layer
smart order router (SOR), which places the child orders into multiple trading venues to
accomplish and tie up the final execution.

As per the 2020 BIS report on FX execution algorithms and market functioning, “Execution
algorithms (EAs) are automated trading programs designed to buy or sell a predefined
amount of securities or FX according to a set of parameters and user instructions. In contrast
to other common types of algorithms such as market-making or opportunistic algorithms,
the sole purpose of EAs is to execute a trade as optimally as possible.”
The report states, “FX EAs came into use more than 10 years ago, and today account for an
estimated 10–20% of global FX spot trading, or approximately USD 200–400 billion in
turnover daily.”
Although not as prevalent in FX as in equity markets, algo trading is slowly catching up, and
it’s only a matter of time before it evolves as a mainstay in global FX. As per the latest
reports, large algo providers and multi-bank platforms have reported consistent increases in
algo volumes over the last few years.

What is Steering the Rise?
The growing adoption of FX EAs in recent years can be attributed to several drivers.
Firstly, the rising electronification of the FX market, especially in FX spots where liquidity can
be accessed via multiple trading platforms.
Regulatory oversight has been another factor driving the adoption of EAs by participants.
Buy-side is more accountable now for how it executes FX trades. “Best execution”
requirements introduced by MiFID II in Europe and elsewhere in various forms led the buyside
to ask for more transparency and automation in the execution process. Although the
best execution requirement exempts FX spot trading as of now, it nevertheless strongly
impacted it. Moreover, the FX Global code of conduct will also drive algo adoption.
The proliferation of multiple trading venues like single bank platforms, multi-bank
platforms, ECNs, direct trading etc., has resulted in the fragmentation of FX liquidity.
Navigating this siloed market is impossible manually; EAs help users bridge the gap and
access, monitor, and execute in the fragmented FX market. Ironically, EAs have also
contributed to market fragmentation, facilitating dealers' internalization of smaller child
orders.

Status Quo
The Covid pandemic and the resulting volatility spike ushered in increased FX EA usage. The
market participants appreciated the robustness and execution outcomes algos provided in
times of high volatility. However, algo adoption in FX has been relatively slower.
The Finance Hive and Bloomberg recently published a report on their analysis of survey
responses from 52 buy-side heads of trading desks. The report states that the US buy-side
executed an average 25% of their flow algorithmically, while their European counterparts
executed 35% of their flow via algos. 36% of the respondents expected the flow to increase
in the next 12 months. The report noted that 56% of participants utilize liquidity seeking or
implementation shortfall algos. Only 13% used the TWAP and VWAP algos, highlighting buyside
bias towards minimizing market impact and reducing slippage.
The buy-side is becoming more demanding, and they are evaluating the performance of
their productive EAs and assessing their providers, ranking and tiering them for future order
flow. Important post-trade performance metrics include fill rate versus benchmark rates,
spread capture, fill venues, speed of execution, and revaluations post-execution.
As the buy-side explores alternate sources of liquidity, execution transparency and easier
algo integration with their tech stack, the algo providers like Banks and independent
vendors are obliging. They are investing in refining the performance of their existing algos
on the one hand while expanding their algo suite on the other. Banks like Citi, ANZ, Barclays,
BNP, etc. have added new algo offerings for their clients. Commerzbank went live recently
with FXall’s Forward First Fixing (FFF) product which intends to reduce cost uncertainty for
algo clients. There has been a recent burst in the number of independent algo providers
providing clients with requisite technological and support tools to execute algorithmically.

Risks
Operational risks that arise due to the failure of algorithms need to be assessed and actively
managed. The providers must thoroughly stress test EAs in simulation environments before
onboarding them in productive systems. Kill switches and other circuit breakers must be
installed to prevent unintentional behavior.
EAs expose users to market risk as opposed to trading at the risk transfer price. The
participants should understand and communicate their roles and capacities (agent,
principal, or hybrid) when trading with one another. A key element is how risks are shared.
In most cases, users take on market risk, whereas providers take credit risk and operational
risk.

The Road Ahead
Adoption of EAs in the future would depend on the execution effectiveness EAs provide to
the users.
Flexibility in EA usage and better user experience with more control mechanisms will be
offered to users at various workflow stages of the algo execution like pre-trade, in-flight and
post-trade.
“Algo wheels” is an upcoming theme in FX that automates the allocation of trades across
various liquidity providers and their EAs and quickly switch from one strategy to another.
These are widely used in equity markets and are now expanding into FX. The algo wheel
usage will bestow workflow efficiency and data-driven results to participants. Moreover, it
naturally fulfills the best execution regulatory requirement of the buy-side.
Electronically traded NDF volumes are rising, making it a potential future growth area for
EAs. Few NDF algo providers offer basic strategies but have started to note considerable
volume increments already. However, NDF algo adoption is still nascent but has much
promise for the future.
Another emerging area is application of AI and ML techniques in the design and
development of EAs. Historic execution data could be leveraged to dynamically adjust the
algo decision parameters based on current market conditions during execution. The next
step will be using post-trade execution metrics to create a feedback loop into the pre-trade
analysis. For example, when a firm executes through an algo and liquidity conditions
change, the data can be fed back into the system to create automated workflows where
trade execution is untouched and exceptions are monitored.

Wrapping it up
Technological innovation has led the FX markets to come a long way from being completely
voice based until a couple of decades ago to a significant chunk being executed
electronically today. Execution algorithms made a crossover from equity markets into FX.
They quickly gained wide acceptance on the back of the benefits offered, like enhanced
automation, reduced market impact, execution transparency and best execution. The next
wave of technology change is set to raise the bar even higher with more sophisticated algo
models unfolding and venture into new growth areas like Algo-wheels and NDF trading. All
these changes are set to spur further adoption and evolution of EAs in FX markets.