FX traders at buy side institutions have embraced algorithmic execution in increasing numbers over the past few years. In part, their decision to use this execution style has been driven by a concerted sales effort that promises tighter execution spreads, lower market impact, and execution intelligence when sophisticated algorithms are employed.
To support this push for algorithmic execution, bank platforms have provided TCA that attempts to measure the effective delivery of the three promises described above. On the surface, all this makes perfect sense and appears to work as promised, but I think still waters run deep when it comes to this positive but complicated trading choice. It is critical to examine the connection between algorithm execution intelligence and the liquidity pools with which each bank platform chooses to interact. This examination will drive a better understanding of how algorithmic execution supports the pursuit of best execution and will identify some questions that the FX trader should ask when choosing an algo provider from a growing list of available suitors.
When an FX trader decides to use an algorithm to execute a trade, in theory he or she has performed a pre-trade market analysis that assesses the current market conditions including real market volatility, the relative strength for the pair being traded, and any skew that may be present in that pair based on trade size. Armed with that intelligence, the FX trader may conclude that using an algorithm is likely to deliver better results than employing a simple risk transfer and paying the spread required to execute the trade size being contemplated.
With the decision made to assume the market risk, the FX trader launches the order and waits for the results. It is exactly at this moment that the algorithm interacts with the liquidity pools that it faces to deliver the hoped for trade outcome. So it seems pretty critical to understand that interaction. But a survey of buy side FX traders would likely reveal that they trust the algo provider to choose the right pools to face or have not bothered to inquire about the characteristics of those liquidity pools. Having made the right decision to use the algorithm, they miss the opportunity to complete their best execution analysis by neglecting this important step.
What Lies Beneath
So let’s dive a little deeper into some concerns that our buy side FX trader might evaluate if he or she chose to examine what lies beneath the still waters. A first step would be to identify the specific liquidity pools that the algorithm faces and understand the characteristics of those pools. After all, the participation of many liquidity pools does not necessarily create more liquidity if they are supported by the same liquidity providers. A key differentiating characteristic would be whether individual pools offer “firm” pricing or allow market makers to employ a last look – the presence of high frequency traders in the liquidity mix adds another issue to be considered.
Another important factor would be understanding whether the algo provider employs internalization. Internalization is typically esteemed in the context of delivering low market impact, but there is certainly more to digest on this topic. A third factor is the use of “mid-pools” and whether the use of such third party markets is appropriate in the context of buy side order flow. A final element is understanding exactly how the algorithm assesses each pool in the context of an order and understanding the financial incentives that might impact which liquidity pool is hit and in which order.
Last look market makers in algo facing liquidity pools are certainly in the position to read market data and take advantage of order flow
Firm vs Last Look Pricing
Algorithmic execution typically searches for the lowest possible spread for each child order which potentially can threaten the even more important buy side goal of low market impact for the entire order. Last look market makers in algo facing liquidity pools are certainly in the position to read market data and take advantage of order flow. When they jump on the trade bandwagon, they can create negative market impact. A buy side FX trader should want to know who can act on his or her algo execution mid-flight.
Use of Internalization
Many algo providers offer their internal liquidity as a guaranteed low market impact alternative. But understanding what comprises this internal liquidity is critically important. Does this internal liquidity face real buy side flows? If it does, this would clearly result in lower market impact. Or does this internal liquidity include flows from the banks’ other trading desks, such as their proprietary desk or options desk, that are running alpha generating FX strategies. In that case, market impact could be negative.
Mid-Pool Presence
Understanding the presence of “mid-pools” is also important. A large FX mid-pool exists that was created to allow banks to offset balance sheet exposures with their bank colleagues. Arguably, buy side order flow should not be included in such a pool since the characteristics of buy side flow are distinct, and the financial incentives of the mid-pool rules would seem to create a conflict of interest for an algo platform using this mid-pool as a liquidity destination for client flow. The Global Code of Conduct is pretty clear when it comes to such conflict of interest issues.
It is the buy side’s obligation to understand all the complexities of this trading style
Rules of Engagement
When the algorithm starts searching for the best indicated price or lowest spread, where does it start? How does it sort through the various pools? Does it consider market impact? Can it measure or model potential market impact? How does it choose a price provider in the event of a tie? Does it consider what’s best for the buy side client or does it break the tie based on what’s best for itself? Since the algo provider is providing a fee for its services, it would seem to make sense that complete transparency would be in the interest of all parties given the conflicts that could occur under these scenarios.
Performance indicators
Once these questions have been asked and answered, there are some key performance indicators that algo customers should use to assess the performance of algo executions:
- Price slippage from first child execution compared to the average executed rate while considering market volatility and other real time factors;
- Execution speed – especially when using passive strategies;
- Trade distribution among liquidity providers – aggregated liquidity is pointless if you trade the majority of your exposure with one liquidity provider;
- Impact of internal pools during the execution workflow.
No doubt, the use of algorithms by the buy side is a good thing and here to stay. That said, it is the buy side’s obligation to understand all the complexities of this trading style. Navigating these waters is not treacherous but exploring what lies beneath the surface is important for buy side institutions who take the principles of best execution seriously.
