Three settings
- Required. A hard rule: a judge never judges outside their categories. A panel that cannot be filled from the right judges stays short, with the reason.
- Preferred. A cost of 25 points per visit outside a judge's categories, weighed against loads and spread. The assigner never leaves a seat empty to avoid it.
- Off. Any judge may judge any project. Useful for a hackathon or a small school fair where every judge sees every kind of project.
Worked example: preferred against required
The sample fair in the judge assigner has six categories and twelve judges with two categories each, plus the rule that no two judges from one school or employer share a panel. With matching preferred, the assigner fills 108 of 108 seats and sends 4 visits to a judge outside their categories.
Switch the same fair to required and it fills 104 of 108, leaving 4 projects short. The first one on the list reads: "P104 Plant growth under LED light: has 2 of 3 judges. Of the 10 other judges, 2 have a conflict with it, 8 do not judge Biology. Only 2 judges can judge it at all under these rules, and the rules ask for 3." The maximum flow check puts the ceiling at 106 seats, a ceiling that ignores the one school rule, so the assigner says the rules might allow up to that many rather than claiming a proof. For this fair, preferred is the better choice: 4 visits outside a category are easier to live with than 4 empty seats.
Expertise tags
Tags are short words a judge enters: soil, robots, water filter. A tag matches a project when every word of it appears in the project's category, flags or title, with plurals folded, so plants matches Plant growth under LED light. Each visit without a match costs 3 points, far less than a category miss, so tags only decide between judges who are otherwise equal. In the sample fair 21 visits match a tag.
Common mistakes
- Requiring category matches with one specialist per category, then finding panels short on the day.
- Letting judges tick every category to be helpful, which hides who the real specialists are.
- Using categories that do not match the ones on the entry forms, so nothing lines up.
- Writing expertise as sentences instead of short words the titles actually use.
- Forgetting that a generalist with no categories ticked can judge anything.
Short panels are explained in how many judges per project, and conflicts in judge conflicts of interest.