The Medallion Fund became the most successful hedge fund in history, but its approach was revolutionary: 100% model-driven trading. No human ever made a trading decision. No one could walk into the trading room and say "buy IBM" or "Google's too high, let's sell."
The strategy was built on finding and exploiting market anomalies. Simons rejected the efficient market hypothesis, which claims prices reflect all available information. "That's just not true," he stated flatly.
Early on, commodities and currencies showed tendencies to trend. Simple momentum strategies could work. But as markets evolved, these obvious patterns disappeared. Renaissance had to find subtler anomalies - patterns that persisted but weren't obvious enough for others to exploit.
The firm ingested terabytes of data daily: prices, volumes, weather, annual reports, quarterly filings, news - everything except "hem lengths." They tested countless predictive schemes, keeping what worked and discarding what didn't.
This was machine learning before the term became popular. The key was having enough data to distinguish real patterns from random noise, then combining multiple weak predictive signals into a powerful ensemble.
The religious adherence to models was crucial. You can't backtest a human's gut feeling, but you can rigorously test an algorithm on historical data.