Why are dark pools a good fit for FX and what lessons can our industry learn from the way they work in other markets?
FX is a highly electronic and fragmented market, where participants trade for a variety of reasons – from transaction-driven and hedging, to speculative trades. Styles range from passive to liquidity-seeking, and even information-seeking (aka high-frequency). The problem for participants with larger orders or those trading less liquid currencies is that the act of trading itself can have an adverse impact on prevailing prices, creating what is termed ‘negative market impact’. Buy-side and sell-side traders face the challenge of minimising market impact and managing market drift.
Mid-matching in dark pools is a great tool to mitigate market impact. It provides the opportunity to cross with other participants who also want to move risk ‘quietly’. In Equities, this is a well-established part of market structure and accounts for 30-40% of institutional trading. Access is typically achieved via algos and may be a combination of private broker/bank-crossing networks and broader access dark pools. In FX there is an opportunity for dark pool trading to account for a comparable market share.
What are the potential benefits of utilising dark pools of liquidity for firms undertaking algorithmic FX trading?
Algo trading can simplify access to dark pools for the buy-side because in effect, the sell-side provider is taking care of both venue connectivity and the order routing framework. Some of the algo banks that Siege works with offer customisation to the client such that they can opt in or out of liquidity sources and, most importantly, rest inactive balances of the parent order in Siege MidPool in the dark. This means that users can curate venues and risk outlets, and also try to maximise the dark pool opportunity while the algo is passively executing in the ‘lit’ market, or even better a bank’s internal pool.
Providing sufficient liquidity is essential for dark pools to gain traction amongst the algo FX trading community. How can that be achieved?
It takes time. Siege has been lucky with some very supportive early adopters on both the buy-side and the sell-side; participants who believe in our approach as well as the evolving market structure. However, there is no short-cut to, or standardisation around, participant onboarding processes, vendor management, information security and documentation. After just 2 years, we have a busy pipeline of existing and onboarding participants. This is vital, because, of course, liquidity begets liquidity. As all these different sources of liquidity reach Siege MidPool, the bank algos automatically start to post more orders for matching, which in turn increases the chance of matching for others, and so the flywheel starts to turn ever faster, allowing participants to match their interests quietly and safely outside of the lit market, for the benefit of all.
How have you gone about making the Siege MidPool a fair and safe environment for algo trading firms?
The first step was to have a strong dialogue with both buy-side and sell-side participants who wanted to join the pool. Our pool is not price-forming and participants match on equal terms at a regulated streaming mid-rate that comes from New Change FX (NCFX). This doesn’t suit all trading styles or time horizons.
Secondly, we don’t publish the orderbook pre- or post-trade. Then we have a layer of minimum order size and live controls to discourage, or even prevent, polling for information. Even though participants do not place orders at prices, they can apply price limits. We also operate circuit-breakers to prevent any matches in dislocated markets. Lastly, each MidPool trade is booked against a central counterparty prime broker, so the participants remain anonymous. This combination of features creates an attractive proposition to trade without market impact.
Why is there no market impact as a result of trading in dark pools like yours?
Not all dark pools are created equal. Some source their benchmark rates internally or from participants. This introduces a moral hazard and even the need to periodically compensate participants for the actions of their matching counterparties, the so-called ‘true-up’ mechanism. As outlined above, we take a different approach from sourcing unbiased benchmark rates to preserving anonymity and curating the pool itself. A ‘shark tank’ is an environment for winners and losers, whereas a dark pool like Siege MidPool is a safe place to exchange risk without adversely affecting the price or your next trade.
What work would banks need to do so their algos can seamlessly operate with Siege MidPool?
Connectivity to Siege MidPool is very standard and in fact easier than for most other venue types because we do not publish a market data feed. The interesting work is to set the sequencing of the MidPool versus other internal and external liquidity sources and then how to apply load balancing in an algo provider’s ‘smart order router’. Entry-level participation would be to fire off clips in phases alongside other venue orders, but a more evolved approach would be to rest part of the inactive order balance to maximise the opportunity for the client to match in the dark without impact.
What are your expectations about how quickly the dark pool model will become firmly established in the algorithmic FX trading space and what factors may influence this?
Interest has grown dramatically over the past couple of years as algo providers look to add more customisation in response to client demand. A key factor is that greater experience with TCA has increased the understanding of the hidden costs of market impact. It is also worth noting that interest is growing with buy-side FX participants who either operate their own algos to execute directly or who are looking to outsource that capability.
Looking forward, in Spot FX we are probably only a few years away from Equity-type market structure percentages for dark pools, with other FX products to follow.