Levels and scales

level.key is an opaque identifier, never an ordinal, even when it looks like a number. Levels are always ordered best first; use level.place and level.of (“2nd of 4”) to read a position, not the key.

scale.kind

kind	levels	example
tier	numbered, 1 best	Solitude 3549003998
rank	strict order, P1..Pn	Phoenix Blacklight 2352059734
group	one level, membership only, no order	Slaughter heart patterns 3659015060
sublevel	named levels without numbers, best first	Solitude “Full White Clouds” over “White Clouds”
band	numeric thresholds on a value	Coco blue percentage (legacy, not yet re-read)

A group scale carries best: null, worst: null — membership says nothing about order. A rank scale always carries bounds, because a strict order has one. scale.kind changes how a UI might draw a system (a numbered ladder vs. a flat badge vs. a named list) — it never changes how a consumer reads one: always via level.place / level.of, never by parsing level.key.

The verdict fields

One seed, in one system, of one guide:

field	meaning
level.key, level.label	the author’s level, exactly as written
level.place, level.of	1-based place in that system’s own list, best first, and the list length
rank	the author’s explicit rank within the level, else null
equivalent_to	the seed the author says this one equals, else null
value	{name, value} the author wrote beside this entry, else null
evidence, quote	text, caption or image, and the verbatim span it came from

{
  "level": {"key": "1", "label": "Rank 1", "place": 2, "of": 4},
  "rank": null,
  "equivalent_to": null,
  "value": null,
  "evidence": "text",
  "quote": "151 168 179 321 387 555 592 617 670 760 809 828 955"
}

A seed listed twice in one system (a recorded duplicate_seed) gets two verdicts. Nothing is deduplicated — the guide is stored as written, contradictions included.