How Long Before a New Book Proves Itself

A new book goes live with a genuinely positive expected return, but daily P&L is noisy — a good strategy can lose money for weeks purely by chance. The question "how long until I can tell this book is actually working, rather than just having a lucky streak?" has a surprisingly demanding answer, because distinguishing a small true edge from zero requires a lot of independent observations relative to how noisy each observation is.

Roughly, the number of days needed to detect a Sharpe ratio $S$ at conventional statistical confidence scales like $(1/S)^2$ times a constant — halving the Sharpe ratio roughly quadruples the time needed to be confident it's real. A strategy with an annual Sharpe of 1.0 might need on the order of one to two years of daily data before its track record clears a standard significance bar; a Sharpe of 0.5 can need four times as long.

For example, a book targeting a modest annual Sharpe of 0.5 with daily returns needs roughly several hundred trading days — well over a year — before the mean daily return is reliably distinguishable from zero at typical confidence levels, even though the book was profitable in expectation from day one.

This is why allocators rarely pull the plug on a new book after a few losing weeks, and why "waiting for statistical proof" before sizing up a strategy can mean waiting through most of its early edge decay.