Eric Stockland, managing director in electronic trading at BMO Capital Markets, said the buy side should ask brokers to experiment with their algo wheels in order to incentivize innovation.

Stockland spoke at TraderForum’s annual member conference on the broker perspective on the implementation of algo wheels.

Algo wheels are an automated routing process which assigns orders to a broker’s algo based on the best fit to a preset list of conditions and have helped the buy side to implement a more systematic approach to trading.

Brokers have to compete for a place in the wheel based on the quality and execution performance of their algorithms.

Leveraging an algo wheel is not an either-or proposition. “Asset managers don’t have to put either all their flow on a wheel or none of their order flow on a wheel,” Stockland said.

One primary reason to use an algo wheel is that it can remove bias in the allocation of orders to brokers, as decisions are based on data. Therefore, the results are more meaningful for asset managers.

“We are all humans and we tend to remember our negative experiences more vividly than our positive ones,” Stockland said. “For example, we may have a bias that a particular broker is good or bad at small-caps.”

However, he warned that removing bias can also be a disadvantage as it is grounded in experience and knowledge. Wheels also remove discretion from the trader, which can be a valuable tool. Therefore, order types that are repetitive and do not require discretion make better candidates for algo wheels.

“We use the crawl, walk, run analogy,” added Stockland. “You don’t have to put the most complicated highest-risk orders in the wheel at the start.”

Another benefit of algo wheels is that they provide contemporaneous feedback on broker performance, and asset managers can see relative performance amongst the group of brokers on their wheel. However, for this process to be most beneficial it is important for all algo providers to have the same set of constraints.

“The buy side gets to dictate the terms, the benchmark and how they want to judge brokers,” Stockland noted.

In addition, algo wheels provide a defensible process for how trades are allocated as performance is systematically evaluated and regulators are increasingly focusing on best execution.

“When auditors come in and ask why so much flow has been sent to Broker A and so little to Broker B, you can point to the algo wheel,” said Stockland.

Algo wheels also provide data to allow the buy side to continuously evaluate brokers and justify why they are increasing or decreasing flow to firms, which have always been hard conversations. In addition, algo wheels can be used to evaluate new brokers.

“Brokers are famous for knocking on doors and telling the buy side that they have reinvented sliced bread,” he added. “Investors need a process to evaluate new and interesting providers.”

Planning Algo Wheels

In order to plan algo wheels, users need to define their benchmark or multiple benchmarks, in advance as objectives will differ for each algo depending on the degree of aggressiveness. Assessment can also use other proxies for performance such as the quality of quote capture rates or the variance of the benchmark.

“One firm might care about quote capture rates, or fade rates, but schedule-driven algos and more neutral liquidity seekers will be different,” added Stockland. “Objectives vary by algo type, so that means buy-side firms probably need a different wheel for each trading style.”

*New* Algos, ATSs and Automation in Equity Markets https://t.co/1V6xQm1UDV via @CoalitionGrnwch by @GreenwichShane

— Coalition Greenwich (@CoalitionGrnwch) April 28, 2021

Brokers prefer as much data as possible enter the wheel so they have more flexibility in segmenting the orders, e.g. between large and small orders, or low and high market cap names.

The three most popular algorithmic trading strategies in 2020 were implementation shortfall, volume weighted average price (VWAP) and dark only according to a survey from consultancy Coalition Greenwich. The report, Algos, ATSs and Automation in Equity Markets, found that institutions valued liquidity most highly in their routing.

Shane Swanson, senior analyst for Coalition Greenwich Market Structure and Technology, said in the report: “In a difficult year, it appears that many firms turned to the tried and true to get their orders executed.”

The use of implementation shortfall rose from 16% in 2018 and 18% in 2019, to 20% in 2020.

“Firms were leaning heavily on their brokers to deliver against the arrival benchmark in a year of dramatic ups and downs,” added Swanson.

Evaluating results

Market orders are always preferable in order to evaluate results and it is usually most useful to look at different durations, for example, looking at orders in the last hour of the day separately from the whole day. In addition, orders can also be grouped by size relative to average daily volume.

“All else being equal, higher participation means higher impact so conflating results is risky,” added Stockland. “Separating out ETFs from single stocks is also important.”

He said normalizing the results by spread is a simple way to get cleaner data, especially if there is a sense of what an order should cost. Stockland gave the example of an evaluation of a wheel for arrival price performance, and how results varied between unfiltered and filtered data.

“We have observed that some investors have the same group of providers in their wheel, who they have known a long time, but they should also provide an opportunity for up-and-coming competitors,” he added.

For example, some buy-side firms have five slots in their wheel where four are given to individual brokers. The fifth is open to five brokers, who each receive one-fifth of the allocation, in order to gather data and test their performance.

“From a broker perspective we love it when clients encourage us to experiment,” said Stockland. “Of course there is a risk of failure but it is one of the ways we find improvements and it incentivizes innovation.”

He continued that brokers also find the feedback on performance relative to peers to be insightful.

“That allows us to triangulate where we are in the stack,” Stockland concluded. “It’s about an incentive structure and transparency in setting up the wheel, communicating the results, understanding the terms and conditions of the competition.”

TraderForum has made a recording of Eric Stockland’s algo wheels presentation available for you to view. You may also submit your information here if you would like to discuss algo wheels further or schedule a meeting with Eric.