Factor Backtest v2 and In-Sample Honesty
What the v2 factor backtest measures, and why its results are labeled in-sample research rather than validated out-of-sample performance.
Category: Factor Research
What v2 simulates
The v2 backtest builds a factor-matched equal-weight benchmark, pure long-short factor benchmark returns, composite factor-loading sleeves, and stock-level risk-parity sleeves, then reports top-10/20/30 buckets net of costs with attribution and institutional-style performance metrics.
Why it is labeled in-sample
The current multi-factor history is too short for genuine walk-forward validation, so every v2 result is published with an explicit in-sample research label. A strong in-sample Sharpe is a hypothesis to retest as out-of-sample data accrues, not a track record.
How to read it
Treat v2 numbers as research evidence the optimizer can rank, not realized performance. The publication gate, not the backtest point estimate, decides whether a sleeve is eligible for agent consumption.