Your strategy failed. Stop moving the goalposts.

The idea worth investigating

Your strategy failed. You changed the rules until it passed. Now you want to trade it.

That is how a research process becomes a permission machine. A losing year gets excluded. A filter removes the worst trades. A different lookback improves the chart. Eventually, the strategy survives because you have stopped allowing it to fail.

A 2014 paper by David H. Bailey, Jonathan M. Borwein, Marcos López de Prado and Qiji Jim Zhu shows how little searching can be enough. In their one-year, independent-trial model, ten strategy variants with zero true Sharpe produce an expected best backtested annualized Sharpe of about 1.57. That attractive number can come entirely from selecting noise.

This is an approximation under idealized assumptions, not a finding that every ten-variant search is worthless. It establishes something more useful: an impressive winner can exist even when the entire opportunity set has no edge.

The expensive mistake comes next: treating the winner as evidence of skill while forgetting the search that produced it. Your effort does not make the result more credible. Every extra attempt gives chance another opportunity to impress you.

This is a research review and editorial argument based on the supplied study, not a new market backtest. We checked the formula arithmetic. The deployment standard below is our judgment, not a rule statistically validated by the paper.

• How a search through losing ideas manufactures an impressive winner.
• Why a reused holdout and a longer history can still leave you exposed.
• A concrete rejection standard for strategies that have not earned capital.

AI assists curation and drafting. The research status distinguishes paper reviews from house tests. Read our research approach.