Judging hackathons at Stanford, MIT, and Harvard in the same stretch of months gave me an unfair sample size. You walk a floor, you hear a hundred pitches in an afternoon, and after the first dozen you stop hearing the idea at all. What you start hearing is whether the team actually built the thing or just narrated it convincingly over a slide.
The teams that stuck with me had one trait in common: they could show me the ugly part. The failing edge case, the thing they ripped out at 3am, the piece they knew was held together with tape. Confidence about the polished demo is cheap. Specificity about what is broken is the real signal, and it is the same signal I look for when I hire.