Huberman uses the same logic with paper selection: if your 10 bad papers on an effect are only 10% likely to replicate that means the effect is 100% likely to be true