Notes
Research Notes
Short, self-contained notes. Each one answers a single question, with the math and a way to check it.
- factors · statistics
Why does rank IC like low-volatility stocks?
On U.S. equities, low volatility and low beta have a rank IC with a t-statistic above 2 and no return spread at all, and size has a negative rank IC with a positive spread. A rank correlation compares typical outcomes, and the typical return of a volatile stock is well below its average.
- backtesting · overfitting
The overfitting statistics said it was fine. It lost money anyway.
A strategy with a deflated Sharpe ratio of 1.000 and a 7% probability of backtest overfitting had a net Sharpe ratio of 1.50 in sample and −0.37 on the five years held back. The statistics were right about what they measure, which is narrower than it sounds.