A Smart Order Router (SOR) is a component within an EMS (Execution Management System) that makes automated execution decisions on behalf of the trader. A SOR uses:
● Order Instructions
The criteria of the order provided such as size, time in force, aggressivity and preferred venues, etc.
● Data (historical data)
Historical data, hit ratios, slippage, and other benchmark information that may influence the execution behavior by directing it toward actions more likely to achieve the desired outcome.
● Data (real-time)
Current quotes and other real-time sources of data such as market prices, volatility, and the depth of both.
● Business Context
Firm based information including positions, profit & loss (P&L), and other executing orders providing netting / internalisation opportunities.
A collection of settings within the SOR that are used to influence its behavior based on other inputs. This policy can be selected using pre-trade rules or automation, also driven by data. For example, a simple policy could contain parameters such as permitted visibility of passive orders, min / max child order size, acceptable slippage, permitted markets, etc.
All these elements are combined by the SOR within microseconds to determine the best action and are not dissimilar from the decision-making process undertaken by a trader. However, in the case of the SOR, the “gut” decision made by a trader is supplemented by the use of real-time data.
Algo or not Algo?
A common question asked is whether a SOR is an algo (algorithm). Technically, the SOR rests upon the same technology used to develop and engage in algorithmic trading. In a MiFID context, it can be argued that distinct from an algorithm, the SOR does not ever fail to execute on available liquidity. It never removes itself from the market or chooses to pass an opportunity to execute in the prospect of better future returns. The SOR is about the best possible immediate action and therefore does not require the same level of scrutiny as a more time-based automated policy. The answer is open to interpretation and we see clients deciding in both directions based on their internal evaluations.
The SOR uses event-based action decisions to achieve an outcome that can be viewed in 3 distinctive phases.
- Aggressive Round- The purpose of the aggressive round is to capture liquidity that is readily available to new orders. When a best execution order is received from a client, market data is obtained from all liquidity pools that list that particular instrument. The smart order router then creates different combinations of “solutions” to execute the available liquidity via one or more aggressive orders. It compares these solutions and chooses the best one based on a set criterion: price / liquidity and preferred venue ranking, their order of application, and input parameters. Once a decision is made on the given solution, the order that constitutes it will be sent to the various markets.
- Passive Round- The purpose of the passive round is to manage the position of the remaining quantity of the parent order. If there is little to no liquidity on the venues within the given limit in the aggressive round, passive orders will be placed on one or more venues or dark pools. The choice of the destination venue(s) is done based on the given parameters, order type(s), time-in-force, and the passive criteria and their order of application.
- Sweeping Round - This is a transitory phase that connects Passive to Aggressive and back to Passive. The purpose of the sweeping round is to hunt liquidity through the rerouting of the remaining passive quantity. The algorithm continuously scans the various markets for compatible opportunities, and once found, will activate sweeping to capture this liquidity.
Based on the data sources and the defined routing, the SOR will continue through these phases until the order is executed, terminated, or expired. Its decisions, inputs, and outputs are all fully traceable and auditable to establish best-execution as well as route further improvements to the SOR itself.
SOR in FX - A very middle-aged affair!
Now that we have addressed what a SOR is and how it operates, we can examine some specific behaviours that bring value to the FX trading desk.
As FX has become increasingly electronified, the actual volumes traded in the interbank market as a whole are decreasing (as per CLS). Many of the large volumes driven by systematic and speculative funds have retrenched and the highly fragmented markets of 2012-2015 have given way to a consolidation of liquidity and a significant rise in Full Amount trading. (Full Amount trading is a social contract between liquidity consumer and market maker that an execution will be sent in its full amount and not split into smaller clips between liquidity providers or overtime).
Although we see a significant increase in the electronification of processes across all asset classes, the market consensus is that protecting your LP relationships is a priority over anonymous ECN / sweepable liquidity. It is in this world of highly electronified and positive-behaviour-driven trading that the new SOR must shine.
The Adoption of SOR is as much a tool to achieve effective workflows and save trader time, as it is to improve the outcome of specific execution.
SOR in FX SWAPS / Multi-Leg orders and Crosses
It is possible to use a SOR to evaluate specific opportunities where a combination of Spot and Forward (FW) transactions may have an overall lower cost to execute than an equivalent Swap or a Broken Date ESP quote might outperform the equivalent IMM of the same date.
The goal of using a SOR is not just to compare Streaming Swaps vs RFQ rates to select the best price, but also to factor in the settlement, clearing costs, as well as any multi-leg execution opportunities that may offer a total lower cost of execution or risk ‘Legging Out’ (when you are left part-way through a multi-leg execution with no liquidity).
Even something as simple as avoiding a give-up charge or choosing to cross through EUR vs GBP vs USD could represent hundreds of thousands in savings over a year for many buy-side or sell-side firms.
Shaping the Trade
In FX, one of the most influential factors in execution outcome is “shaping” the trade. Even in a mono-market execution, deciding how to split the clips to reduce impact or slippage, or even whether the full-amount will achieve a better outcome is all data-based decision making that can be performed by the SOR.
Old SOR technology could often not be trusted to respect the Full Amount restrictions and so execution decisions were often completely siloed to avoid Sweeping Full Amount for example.
The new data-driven intelligent SOR can predict market movement, reject rates, last look behaviours, and respect LP relationships by adhering to constraints that protect your LPs from behaviors potentially viewed as toxic as well as mix different liquidity pools respecting rules for how to interact with them.
Examples of new SOR behaviour include:
Double Hit Protection
Ensuring that following any execution, traders are prevented from executing again for an agreed time window. This can vary based on the chosen LP, execution method, CCY, and market conditions. For instance, an emerging market (EM) CCY during low volatility could have a 45 second hit protection window allowing the LP to hedge their risk before the client can trade again.
Micro Volatility Prediction
Using AI/ML agents to predict volatility in a market over the next 100ms. This allows passive pegging to achieve 5-10 pip improvements (on average) in execution improvement by being consistently better placed in the market to capture the largest offset.
Using historical data analysis to evaluate the likelihood of slippage or iceberg liquidity via different execution methods. A probability can then be used to determine if Full Amount, Aggressive Sweepable, or Passive liquidity will give a better overall execution in current market conditions, rather than just making a simple price-based decision on where to execute.
Streams may be prioritised by the SOR by evaluating:
- Accept / Reject Symmetry: Will trade be accepted or rejected in different market conditions symmetrically and can this behaviour be predicted to avoid rejection? Will adverse hold times impact other benchmarking metrics?
- Cost of Rejects: Ultimately, can the likelihood of having an order rejected be predicted, and can the ongoing cost of that rejection be measured? This allows traders to avoid the LP in specific conditions or remove them altogether where the cost of one rejected trade could outweigh the benefits of using them.
- Hold Times: Analysing the last look and hold times between placing trades and fills / rejects which could impact the overall execution quality of the trade, while also evaluating if including an LP in an aggregation of liquidity impacts the overall spread / execution analytics and decay faced while executing.
The powerful range of capabilities allows the trader to specify a more generic outcome, but allows the SOR to, within defined constraints, work out the very best way to achieve it.
ESP vs RFQ
An increase in the availability of Executable Streaming Prices (ESP) streams have also opened new territory to take advantage of SOR technology. ESP streams are no longer the exclusive domain of spot FX but are now regularly offered for Forwards, Swaps, NDFs, and even broker dated FW / Swaps. A trader can now send an IMM date to the SOR that is collecting data on thousands of real-time prices each second to execute with the best LP, rather than the traditional RFQ alternative limiting selection to a small panel of LPs while leaking intent to the market.
Multi-Market and ECN
Once the scope is expanded to executing across counterparties, as well as introducing passive opportunities, operating without a SOR becomes as dangerous as it is foolish. The SOR’s microsecond decisions account for market movements, passive rebalancing, gaming protection, ghost liquidity, and High-Frequency Trader (HFT) participants. Without a SOR a firm’s participation in these markets simply becomes fodder for everyone else to profit from.
Next Generation of SOR - Data Intelligence
The evolution of SOR in FX has been the move from high-performance real-time decision making, to multi- workflow automated trading. However, the selection of these workflows are still highly limited to determinations made by the traders themselves. Configuring a SOR to decide between which ESP venues to execute what clip size on has evolved to a total decision between multiple streams of liquidity with different behaviours and rules.
The future evolution of this is to provide real-time feedback to the trader on workflows they may not have even considered through simulated decisions on millions of combinations of liquidity and execution methods: A true tool for smart order workflows and not just smart order routing. It is a paradigm shift that turns the concept of SOR on its head.
FX etrading has already automated a great number of routine jobs. The yesteryear trading floors with hundreds of traders on the phone are today replaced by a few silent traders manning exceptions in the systems that process in milli-to-microseconds tens of thousands of quotes and orders each second. In reality, machine learning (ML) will bring intelligent decision making to systems which for a very long time aimed to only become faster, but not always smarter.
ML will allow near time predictions and analysis which will drive systems and decisions. It means that the silent trader will now have a better understanding of what the system and overall market is doing. Any company not investing today in this area will be at the risk of not being competitive in the near future.