Pragma launched Mercury, its deep neural network-based execution engine, in 2021 after a multi-year research and development effort, and a year of controlled production trading trials with its clients. Mercury’s deep neural network microtrading engine controls the routing, sizing, pricing and timing of orders, and has shown a robust improvement in execution quality, with an average shortfall improvement ranging from 33% to 50% across billions of traded shares.

Support for Canadian and interlisted equity trading is a natural extension of Pragma’s deep learning capabilities. Leveraging its high-quality simulator, a decade of experience in trading interlisted equities, and an additional year of research to extend the neural-network model to the problem of interlisted trading, Pragma clients can now access more intelligent execution algorithms for Canadian and interlisted equities. Through Panorama, Pragma’s algorithmic management system, clients can also view customized A/B experiments with the new Mercury algorithms on a real-time basis.

“Efficient interlisted trading is a very tricky problem,” said David Mechner, CEO of Pragma. “A trader has to deal with different rules and market structure, as well as significant fragmentation across exchanges and alternative venues on each side of the border, and an additional FX component. This complexity was an R&D challenge, but ultimately plays to the strength of the Mercury deep neural network microtrading engine, which learns through training to exploit complex multi-dimensional interrelationships across stock characteristics, venue characteristics, the liquidity demands of the order being traded, and real-time high-frequency market signals.”