Among barely-significant wins from small tests, the true effect is ~50% smaller than reported.

Among barely-significant wins from small tests, the true effect is ~50% smaller than reported.
The winner's curse / Type-M (magnitude) error: when power is low, the only effects that clear significance are the ones noise inflated. You ship the exaggeration.

— Low-power + significant = expect regression to the mean on rollout.
— Bigger samples don't just confirm wins; they right-size them.

Reported ▮▮▮▮▮▮ vs Replicated ▮▮▮.

Read the number, not the story. [n