Banks’ machine learning/AI-based algorithms are gaining traction and generating alpha
New research published by TABB Group says the sell-side is in a “survival of the fittest” race for the top spot in buy-side clients’ algo wheels.
According to New York-based senior equity analyst Michael Mollemans – the author of “AI in Sell-Side Equity Algorithms: Survival of the Fittest,” –the sell-side’s artificial intelligence (AI) equity algorithm ecosystem has expanded after years of development work to a point where significant AI-attributable excess returns have finally begun to be realised in the past two years.
As automated performance measurement applications like algo wheels are driving broker selection decisions, competition to build better, faster, smarter algorithms has become a war of attrition. “Now more than ever,” says Mollemans, “not keeping up means you’re going backwards, which is why we believe consolidation in the algorithmic trading space will continue, just as the sell-side overall continues to consolidate.”
TABB Group interviewed 50 AI algorithm experts from the buy side, sell side, and fintech vendors and produced AI algo ecosystem case studies on US, European, and Asian banks. The 27 page, nine-exhibit report, created to help traders gain depth and breadth of insight and a better understanding about what’s happening “under the hood” in their AI algorithms, covers eight key areas:
• How sell-side firms must stay ahead of rapidly evolving, AI-algo data science
• Improvements in performance attributable to AI models
• Leveraging economies of scale and development budgets to support advanced AI ecosystems
• AI applications focusing on scheduling, price and volume prediction, spread capture, strategy and parameter selection and venue-routing decisions
• “Explainable AI”
• Turning a “black box” algo into a “clear box”
• Utilising proprietary data unavailable to competitors
• How oversight and governance procedures will become more sophisticated
Moving forward, Mollemans believes that only a few banks will dominate the global algorithmic trading space in the next five years.
“The most significant challenge is the changing science of AI and the growing investment needed to transition from traditional algos to AI," he says. "Some of these AI-based techniques, like t-SNE, were not even in existence 10 years ago. In fact, 41 per cent of sell-side firms interviewed launched their client-based AI algos only last year.”