hackathon-judging

Compute judges per team as k = floor((J × W) / (N × t)) from team count, window minutes, and per-team demo time. Full coverage requires N ≤ W / t and adding judges cannot extend that ceiling, since every judge must still traverse all teams.

Weight the categories already named in the published brief using integer percentages summing to 100, keep criteria at four or fewer, and score each on a 1-5 integer scale. Write the weights and scale down before anyone scores.

Normalize only when each judge scored a proper subset of the field, since a harsh judge's constant then lands on one slice and skews the rank. Under full coverage the constant hits every team equally, so normalization is pure overhead.

Publish a deterministic ladder before submissions close: highest score on the highest-weighted criterion, then count of judges ranking the project first, then a named tie-break judge on the record. Avoid coin flips and re-scoring finished demos.

Yes, when a signed agreement requires it: confine the sponsor judge to that track, pair them with a neutral judge, run a drop-test with and without their scores, and disclose the composition in the published rules before submissions close.

Use triage only when N exceeds (J × W) / t, meaning not even one judge per team fits the window. The pre-screen is a published pass/fail gate on submission requirements, and the finals slate must pass the feasibility check on the remaining window.