The Case Against Your Favourite Backtest

Overview

A methodical conversation about the gap between a factor that replicates and a factor you can actually run money in. The guest separates three failure modes — data-snooping, capacity, and the trading cost that only appears at size — and argues the third is where most of the decay happens. The last section is practical: what to demand from a research note before allocating to it.

Key takeaways

• Replication is a low bar; implementation shortfall at size is where published edges die.

• Capacity is a property of the trade, not the signal — the same factor is fine at $50m and gone at $5bn.

• Ask for turnover and realised spread, not Sharpe, when evaluating a strategy.

• Most "regime change" in factor returns is crowding showing up with a lag.

Companies & tickers

• MSCI
• AQR Capital Management

Topics

• Factor investing
• Quant
• Research methodology
• Trading costs
• Capacity