Why our numbers are right (and chatbots' aren't)
“Mastering ESO” isn't a knowledge problem — it's a freshness-and-exactness problem. This page is the full discipline we use to solve it, so you can decide for yourself whether to trust us.
Start a free build on this data
The data spine: datamined, not remembered
No number on this site comes from an AI. Every value is datamined game data pulled via UESP's esolog — the community's established datamine of the actual game client. Right now that spine covers 712 sets, 541 skill morphs across 70 skill lines, the full champion point star tree, per-skill damage coefficients, and the achievement catalog — every page stamped with its pull date (2026-08-23 right now).
We don't editorialize values. Bonus text is shown exactly as mined. Where the datamine has no category for a set, we say “Uncategorized” instead of guessing; where a value is missing, we show nothing rather than a plausible-looking fill. The buff finder is built from the same mined effect identifiers, not forum lore, which is how it can say definitively which sources of a buff are wasted overlap.
The refresh discipline: freshness is the design
Patches trigger re-pulls, and quiet weeks change nothing. A scheduled job re-runs all six datamine ingests weekly. If a pull comes back identical, the committed data is kept as-is — no churn, no fake freshness. If content actually changed, the new pull ships and every affected page updates in one run, because the whole site is re-rendered from the data spine.
Bad pulls are rejected automatically. An ingest that fails, or returns a file more than 20% smaller than the last good one, is rolled back rather than published. The failure mode we engineer against is the silent one: nothing breaks when data goes stale, the pages just quietly start lying with a confident date on them. The stamp you see is therefore a real guarantee — it can never drift more than 14 days behind a genuine re-pull.
Where the live meta comes from
The measured meta is the one layer not mined from the client: it's aggregated from top parses on ESO Logs — currently 19 endgame encounters, snapshot dated 2026-09-05. We report class dominance, not individual players' gear: ESO Logs only exposes per-parse gear when the uploader logged it, and we don't publish numbers we can't stand behind. Top-parse data skews toward optimized raid groups — read it as “what's strong”, not a rule for casual play.
What this site is not
We're an unofficial fan resource — not affiliated with ZeniMax Online Studios or Bethesda. Datamined values can differ from in-game tooltips in edge cases, and PTS values change before going live, so verify in game before expensive decisions (gold, crowns, or transmute crystals). The legal version of that promise is on the disclaimer; what we collect about you (very little) is on the privacy page.
FAQ — frequently asked questions
How often is ESO Decoded's data refreshed?
A scheduled job re-runs every datamine ingest — sets, skills, champion points, skill coefficients, achievements, and the live meta — weekly, and patches trigger extra re-pulls. If a pull returns nothing new the data is left untouched, and the visible retrieval stamp is never allowed to drift more than 14 days behind a genuine re-pull. The current spine was retrieved 2026-08-23.
Why does a number here differ from a popular build guide?
Static guides are written once and decay silently as ESO rebalances. Our values are re-derived from the current datamine and dated, so when two sources disagree, compare their dates — the fresher pull usually wins. If the decision is expensive, verify the tooltip in game.
Why not just ask an AI chatbot for set bonuses?
Language models answer from training-time memory, and ESO changes set and skill values several times a year — a chatbot will quote last year's tooltip with full confidence. Nothing on this site is generated from AI memory: every bonus line is datamined game data with a retrieval date.
What happens when a datamine pull looks wrong?
It gets rejected, not published. A refreshed data file that shrinks by more than 20% is treated as a failed pull and rolled back to the last good version, so a bad scrape can never silently blank out half the site behind a fresh-looking date stamp.