Based on these experiences, we try to formalize the notion of publication bias.

We formalize the notion of "relevance" and propose an evaluation methodology in which statistical significance is retained for relevant improvements, but not for irrelevant ones.