Reflections on Machine Learning in Algorithmic Trading
Recently, I had the opportunity to dive into Machine Learning for Algorithmic Trading (Second Edition) by Stefan Jansen. It’s a comprehensive read, though there’s certainly room for improvement. One key takeaway I resonate with is the author’s assertion: machine learning alone isn’t sufficient in algorithmic trading. Expert judgment remains irreplaceable.
Personally speaking, I think AI/ML in trading is still primarily an assistive tool, demanding close governance and oversight. Unlike other sectors where AI/ML adoption has matured, trading appears to lag behind. Why? Here are a few thoughts:
- Reliance on Traditional Methods: Many financial quants still prioritize traditional approaches rooted in calculus and classical mathematical practices, over newer, advanced techniques like machine learning.
- Scarcity of Cross-Domain Expertise: Success in modern quant trading requires an intricate combination of skills — expertise in AI/ML, mathematics, big data infrastructure, financial markets, global economics, politics, and risk analysis among few others. Talents with this blend are rare but not unattainable.
However, the rise of Generative AI brings hope. This technology acts as an “expert multiplier,” enabling professionals to tackle challenges as if supported by a team of specialists. Imagine the possibilities when seasoned AI/ML practitioners with strategic leadership, financial knowledge, and strategic acumen — grounded in both academic rigor and industrial experience — enter the trading world. These are the professionals who could redefine success in trading and asset management.
That said, we need more talent with strong AI/ML backgrounds in this sector. It’s true — finance can be stressful, and ethical conflicts sometimes push individuals away. However, to maintain or achieve global leadership, engaging the brightest and most tech-savvy minds in finance is crucial.
