How to read our data flags
Published · skill data as of 2026-10-26 · sources & flags
Every tool on this site runs on one rule: missing is marked, never guessed. When a number comes from the 1.0 server files but no source names the stat it modifies, the node says so. When a description is still the pre-1.0 text, the row says so. Those markers are the flag system — this piece is the reader's manual, with the live counts and the stories behind the first flag ever cleared.
The canonical list lives on the data sources page, re-counted from the data on every site build. The numbers below are that table, plus one flag from the armor database that renders as not recorded instead of a chip.
The flags, by how much they should worry you
Trust the graph, weigh the text. Ranks, costs, connections and positions come from the same server files the game runs on — none of those carry flags. The flags concentrate on values' meaning, descriptions and names:
| Flag | Live count | What it means when you see it |
|---|---|---|
| Pre-1.0 vs 1.0 value diff not yet verified | 207 | Carried-over skill; whether its value changed at 1.0 is unverified — treat listed values as current, not as a before/after claim |
| 1.0 value shown; the stat it applies to is not named in the data | 122 | The number is from the files; which stat it modifies is inference-free display, not a claim |
| Values from 1.0 server files; description is pre-1.0 text | 112 | Numbers are live, the wording is the older text (1.0 shipped no description strings) |
| Description is the pre-1.0 text | 102 | Same caveat, description-only |
| Cost per rank assumed equal to the node's listed cost | 91 | One cost is listed per node; we assume extra ranks cost the same until in-game confirmation |
| No description in the data yet | 16 | The node exists with values; no source publishes text for it |
| Name from a single source | 6 | The display name rests on one source — flag clears when a second confirms |
| Attribute effect unverified | 6 | Attribute node whose exact effect text isn't in the data |
| 1.0 description verified in-game | 1 | The good flag: checked against the live game and replaced |
| No published name | 1 | The Unknown node — internal ID 1569371409, mage area |
In the armor database the same idea renders as not recorded: 88 of 340 pieces have a sourced acquisition, and the other 252 show nothing rather than a plausible-sounding location. A blank cell is data honesty, not a gap you should fill from memory.
The first flag ever cleared
In October 2026 a player reported that Warrior's Path displayed 0%/5%/10% with its pre-1.0 description, while the live game shows 5%/10%/15% for one-handed axes, clubs and swords. Two things were wrong at once: the description was stale and a pipeline bug — per-rank values computed as base + per × (rank − 1) instead of the game's base + per × rank — had every one of the 91 multi-rank skills displaying one rank low.
The fix shipped both corrections and the machinery to keep them: Warrior's Path now carries the verified 1.0 description ("Damage with One-Handed Axes, Clubs and Swords is increased by 5% per rank.") under the first descVerified10 flag, ten more nodes were spot-checked against external sources, and every build since re-checks all values against a committed baseline snapshot — a value can no longer silently drift. The full incident log is on the updates page.
Labels with nothing to flag
The flag vocabulary is larger than the live table: labels like Value unverified, Name from patch notes, Pending verification, Sources disagree on the name and Name unverified are defined in the system and currently sit at zero rows. That is deliberate — the vocabulary was written for every failure mode the pipeline can hit (a name-only-in-patch-notes, two sources contradicting each other), so when one appears it gets an exact label instead of a bespoke excuse. An empty label is a clean bill for that category, not an unfinished row.
How a flag clears
Each flag has a specific exit condition, not a vague "someday": a single-source name clears when a second independent source confirms it; a pre-1.0 description clears when someone checks the skill in the live game and the verified text replaces the old wording under the verified in-game flag, as happened with Warrior's Path; an armor not recorded cell fills when the harvest re-runs against a wiki page that now publishes the acquisition. Value flags like the 91 per-rank cost assumptions clear against in-game observation, and any fix ships with a regression check so the same flag can't silently come back.
Pre-1.0 description text carries one more layer worth knowing: it is community-wiki content used under CC BY-SA, and the flag wall between "values from the server files" and "wording from the wiki" is exactly what keeps our claims attributable. The data sources page spells out which source owns which column — names, values, descriptions, item stats — so every flag traces back to a source boundary, not to uncertainty about our own pipeline.
Reading flags in practice
Three habits get you the most out of the system. One: when a guide and a flagged table disagree, the flag tells you which side is unverified — that is the disagreement's origin, not an error. Two: a flagged number is still a number from the game files; use it, just don't quote it as a verified before/after diff. Three: flags are invitations — if you can check something against your own game, the contact page takes reports, and a confirmed correction typically goes live within a couple of days.
The method behind all of it — which source owns names, values, descriptions and item stats, and what each flag commits to — is documented in full on the data sources page, and the short version of how to check anything we publish is now also in the about page. When a flag count in the table above moves, it moved because the data did.