Leigh Drogen - Quant vs Traditional Investors and How Alphas Become Betas - [Invest Like the Best, EP.41]

Episode Summary

Executive Summary: Patrick O'Shaughnessy interviews Lee Drogan, founder of Estimize, about how quantitative investing has reshaped markets. They trace the decay of old sell-side and discretionary edges, the rise of crowdsourced and alternative data, and why the future likely belongs to hybrid “systemental” firms combining machine rigor with human judgment.

Main Topics: From discretionary to quantitative markets (Priority: 5/5): Drogan explains how alpha has shifted from stock-picking toward models, data, and systematic research. How Estimize works (Priority: 5/5): Estimize crowdsources earnings and macro estimates to create a larger, less biased consensus dataset. Research process and overfitting (Priority: 5/5): He stresses ex-ante hypotheses, out-of-sample testing, and avoiding the trap of data-mining. Arms race in alternative data (Priority: 4/5): New datasets like location, satellite, and insurance registrations quickly become table stakes. Hybrid human-plus-machine investing (Priority: 5/5): Drogan argues domain expertise can improve model outputs, especially in sector-specific situations. Behavioral discipline and decision systems (Priority: 4/5): He says many discretionary managers fail because they do not actually follow their own stated process. Market structure and future volatility (Priority: 4/5): More systematic trading may reduce some dislocations, while passive flows raise questions about future crashes.

Key Arguments: Alpha is getting harder; informational edges are increasingly arbitraged away by large platform firms. Good quant research starts with an ex-ante hypothesis, not blind regression fishing. Out-of-sample validation is essential; otherwise backtests overfit and fail in production. Discretionary managers often ignore their own decision rubric, which destroys consistency. Crowdsourcing estimates broadens panels, reduces bias, and can beat Street consensus. Hybrid models can outperform pure quants when sector expertise helps weight signals correctly. Alternative data creates a moving frontier: today’s edge becomes tomorrow’s table stakes.

Data Points: Estimize coverage: about 2,100 companies - Companies covered with three or more estimates Estimize deeper coverage: 1,400 companies with 10 plus estimates - Scale of the crowdsourced estimate panel Apple estimates: 1,000 estimates - Example of a very large-cap name on Estimize Accuracy vs Street: 70% of the time - Estimize consensus more accurate than the Street in aggregate Accuracy in mid/small caps: 75%, 80% of the time - Estimize advantage is strongest where Street coverage is thin IBIS arb timeline: 30 years - Drogan says the earnings-estimate dataset took decades to arbitrage away Launch timing: January of 12 - Estimize was started in January 2012 Critical mass point: mid-14 - Adoption accelerated after the platform had enough data and awareness Data company scale: $15, $20 million in revenue a year - Typical revenue level where many niche data companies sell Passive market share: 30, 35, 40 percent - Approximate share of assets in passive strategies discussed in the interview Flash crash reference: May 6, 2010 - The memorable market event Drogan discussed as a career day Long-only drawdown rule: 200-day moving average - Example of a simple trend filter he suggests for passive portfolios

Pivotal Quotes: "we basically attempted to normalize everything and then basically Z-score the entire universe and go along the top deciles and short the bottom deciles" — Lee Drogan: Describing the core stat-arb method used in his early hedge-fund career "if you don't know why it works, you won't know why it stops working" — Lee Drogan: Why ex-ante hypotheses matter in quantitative research "the thing that I don't really understand" — Patrick O'Shaughnessy: Reacting to Numeri's opaque, crowdsourced machine-learning setup

Implications: Investors should expect faster data decay, greater model competition, and a premium on process discipline; the unresolved question is how much human judgment should remain in the loop.