Validation · 2026-07-14
Deflated Sharpe and the trial budget
The observed Sharpe, discounted for every trial burned in finding it. Bailey and López de Prado, not folklore.
Search enough parameter space and you will find a Sharpe. The industry is full of them. We discount the observed ratio for the number of trials consumed in finding it — deflated Sharpe after Bailey and López de Prado (2014).
The gate is 0.95. It is an internal standard, not a guarantee of future results, and the passed statuses we print are the shape of the mechanic, not a medal for a single candidate.
If the idea only exists at one fitted point, it does not exist. The neighbourhood has to pay as well.