The search for one perfect EA asks too much of one return driver. Every edge encounters regimes in which its premise is weak, execution changes, or losses cluster.
MSRLab instead asks what role a supported edge can play. A modest system may be valuable when its instruments, horizon, and failure modes complement the rest of the book. A spectacular backtest may add little when it duplicates existing exposure.
Portfolio usefulness never excuses a weak hypothesis. Each sleeve must first survive its own evidence gates; diversification is assessed after that, not used to conceal fragility.