Algorithm wheels are gaining significant adoption, with investment firms rapidly implementing new technologies to improve performance and support best execution efforts. However, algo wheels are not suitable for everyone. To help assess best-in-class algo wheel configurations, we’ve developed an evaluation framework.
An algo wheel is an automated routing process which assigns a broker algo to orders from a pre-configured list of algo solutions. ‘Algo wheel’ is a broad term, encompassing fully automated solutions to mostly trader-directed flow. Investment firms typically use algo wheels for two reasons: first, to achieve performance gains from improved execution quality; second, to gain workflow efficiency from automating small order flow or normalizing broker algos into standardized naming conventions.
Algo wheels are on the rise as the buy-side increasingly grapples with the complexity of achieving best execution and the challenge of sourcing liquidity. The multitude of algo offerings, with non-standard names and unique characteristics, can be difficult to navigate and manage, so a handful of firms have leveraged algo wheels to help normalize offerings and manage commissions.
In parallel, the Markets in Financial Instruments Directive II (MiFID II)’s best execution requirement has prompted investment firms to justify broker selection more transparently. As a result, technology vendors rushed in to sell broker-selection wheels and transaction cost analysis (TCA) to help address this requirement. As technology products helped to lower the cost of implementing an automated trading wheel, buy-side use of algo wheels skyrocketed and suddenly the term “algo wheel” became ubiquitous.
To help assess best-in-class algo wheel configurations, we’ve developed the following evaluation framework that outlines steps for your consideration, based on our observations, educated opinions, and experience, having assisted clients during the vendor selection and implementation processes.
Consider whether to route orders into a rules-based, automated algo wheel or to a trader for manual execution.
The expected cost, order size and alpha (proxied by Portfolio Manager/Fund) are among the most popular inputs to manage automated flow segmentation rules in an algo wheel. Additionally, it is important to monitor in-progress orders for algo wheel applicability. Large price movements, poor realized/unrealized performance or deteriorating market conditions may be reasons to remove an order from an algo wheel and actively manage it by a trading desk.
Select the right composition of brokers to participate on the algo wheel.
Consider brokers’ algo platform flexibility and commitment to customization in initial broker selection. It may be nest to remove brokers that both consistently and significantly underperform peers, but only after a grace period to allow for experimentation. Adding new brokers on a trial basis allows you to explore new approaches.
Define a strategy, benchmark and algo parameters at the outset with each broker.
Codify a limited number of normalized trading strategies and define each strategy’s goals, constraints and flow characteristics. Then communicate the details to each broker on the algo wheel and rely upon brokers to optimize their own strategies to best meet individual execution requirements.
Use randomization to remove trader bias in broker selection.
While many firms use a simple coin flip, a modeled approach aims to randomize broker selection while simultaneously increasing the probability that brokers’ sample sets will be similar to each other. This type of approach can reduce the number of orders (sample size) needed to gain insights into algo rank, and therefore be quite valuable to the algo wheel process.
Leverage models to get the most out of your data for broker ranking and actionable insights.
A good performance decomposition model is critical to both remove noise from the broker ranking and to assist in identifying areas of improvement in broker algo behavior.
Optimize allocation weightings.
Most firms currently eyeball how much to increase the allocation weighting for better-ranked brokers, and how much to decrease allocation for poorer-ranked brokers. However, arguably a more systematic and data-driven approach towards allocation adjustments could serve as one of the quickest and least expensive ways to improve the algo wheel experience. A formal, mathematical optimization model is typically quick to write and provides optimized allocation weightings each quarter to inform the trading desk’s decision making.
For institutional investors only. Not for retail customer use.