Evolution of Machine Learning: Ensembles and Boosting

In this episode of Odds on Open, Ethan Kho sits down with Joe Mezrich, Founder of Metafoura LLC and former Managing Director at Nomura Quant Strategies, to reflect on nearly 40 years in quant finance. Joe’s career spans the early days at Salomon Brothers—where he helped pioneer factor models, risk modeling, and even early machine learning in finance—through senior sell-side research roles at Morgan Stanley, UBS, and Nomura.Joe shares how the sell side effectively built modern factor investing, why models like the Barra risk model failed in crises such as the Tech Bubble (2000) and the Quant Crisis (2007–08), and how market-neutral strategies and algorithmic trading continue to shape today’s buy-side. He also explains why interpretability, from CART decision trees to today’s LLMs for trading, is critical for robust risk management.We cover:- Origins of quant finance on the sell side at Salomon Brothers.- Early factor models, the Barra risk model, and portfolio risk modeling.- Use of robust statistics and CART decision trees in machine learning for finance.- Why risk models failed in the Tech Bubble (2000) and Quant Crisis (2007–08).- Growth of market-neutral strategies and interaction between sell-side research and the buy side.- Crisis lessons: liquidity concentration, model speed, and explainability.- Evolution of factor investing into overlays and ETFs.- How quant researchers balance complexity vs. interpretability with LLMs for trading.- Role of alternative data, point-in-time datasets, and data visualization in alpha.- Wall Street culture: Liar’s Poker-era Salomon, Morgan Stanley, UBS, Nomura.- Impact of interest rates, earnings vs. sales growth, and macro regimes on factors.- Sustainability of multi-manager pod shops (Citadel, Millennium) and implications for quants.- Career lessons: curiosity, humility, and finding beauty in quant models.Whether you’re a quant researcher, an aspiring algorithmic trading professional, or an allocator seeking to understand systematic funds, give this a listen.00:00 Intro and Episode Overview00:46 Origins of Quant Finance at Salomon Brothers02:56 Early Factor Models and Barra Risk Model05:51 Robust Statistics and CART Decision Trees08:58 Machine Learning in Finance 1990s Experiments12:06 Why Risk Models Failed in Tech Bubble15:31 Lessons from the 2007 Quant Crisis18:51 Rise of Market Neutral and Sell-Side Research22:26 Evolution of Factor Investing to ETFs26:01 Balancing Complexity and Explainability for Quants29:16 Alternative Data and Point-in-Time Datasets32:46 Wall Street Culture Salomon Morgan UBS Nomura38:08 Interest Rates Macro Regimes and Factor Drivers41:51 Are Multi-Manager Pod Shops Sustainable?46:04 What Makes Exceptional Quant Researchers Last49:26 Curiosity Humility and Risk Management52:56 Finding Beauty in Quant Models and Data56:16 Final Lessons from 40 Years in Quant Finance