Urd Atlas methodology

Methodology

How Urd Atlas turns daily blockchain observations into deterministic network-state reference data: observations, transforms, labels, confidence, scorecards, drivers and traceable JSON artifacts.

What this methodology is for

Urd Atlas is a descriptive network-state layer. The methodology explains how daily chain observations become stable JSON rows that can be read, joined, inspected and reproduced across product surfaces. It does not convert chain conditions into instructions or future-state guarantees.

Read these first

First time here? Start with Public Methodology Reference to understand what the reference data means, then read Validation for the empirical consistency and robustness evidence. Use Verification & Evidence to inspect the published evidence path and Field Dictionary when you need exact field definitions.

Current methodology note: Confidence v3

Current Meta rows use confidence_v3_l2_capacity_required. Confidence retains sqrt(data_quality_score × label_confidence_score) and the0.40 publish gate. Data quality is chain-profile-aware; current L2 confidence explicitly requires the published capacity-utilization evidence used by the L2 ruleset. Confidence is a reliability score for the published row, not a calibrated probability that a label is true.

Product boundary

Methodology pages document descriptive on-chain reference data only: no price data, no forecasts, and no recommendation outputs. Confidence and coverage explain how well the published row is supported; they do not turn regime labels into advice, automation rules or future outcomes.

How to read this section

The methodology section is split into separate pages so that customers can move from a fast overview into more technical detail without reading one single monolithic document.

Page	Purpose
Reference	Canonical public methodology and interpretation rules.
Field Dictionary	Field-level definitions and warnings, including current Confidence v3 fields.
Validation	Empirical internal-consistency, robustness, and current diagnostic evidence.
Verification	Worked examples and evidence path.
Freshness	Publication lag and freshness policy.
Boundaries	What the public methodology discloses and does not disclose.
Integrity	Determinism, row identity, and archival traceability.
Changelog	Methodology updates, historical republish notes, and subscriber action guidance.

The method in one pass

Urd Atlas starts with daily chain observations, applies deterministic transforms, assigns a network-state label, computes confidence, and publishes versioned JSON artifacts. The label is the readable compression. The scores, confidence fields, dates and methodology identifiers are the machine-readable audit surface around that label.

Step	Question answered	Published surface
Observe	What did the chain look like on this observation date?	`Gold`
Transform	What recent and historical context should be compared deterministically?	`Derived`
Classify	Which descriptive network-state regime best fits the evidence?	`Meta`
Explain	What should a human reader know before using the row?	`Briefs`

Reference layer model

Urd Atlas publishes four JSON layers: Gold, Derived, Meta, and Briefs. Gold is the daily observation layer. Derived is the deterministic trend layer built from Gold. Meta is the analytical layer that publishes regime, confidence, scorecard state, drivers, and presentation-ready summaries. Briefs are the readable JSON layer built from the latest Meta context.

The most important page for a technical customer is the public methodology reference. The most important page for an auditor or quant reviewer is the verification page. The changelog should be checked when a customer compares cached historical JSON with newly pulled rows.

Confidence and status model

Confidence answers one narrow question: how well-supported is the current published analytical state by the data and by the label-specific evidence? It is not a probability of a future chain state, a recommendation signal, or a reason to automate a downstream action by itself.

Part	Plain meaning	How to use it
`data_quality_score`	Do the relevant chain-specific fields exist, remain fresh, and cover enough recent/history rows?	Use it to avoid over-reading weak or sparse rows.
`label_confidence_score`	Does the evidence clearly support the specific label that was assigned?	Use it to separate clear labels from ambiguous labels.
`confidence_score`	Composite confidence: `sqrt(data_quality_score × label_confidence_score)`.	Use it as the default gate in analysis and reporting.
`status.one_liner`	Readable explanation that distinguishes the published regime label from adjacent scorecard pressure.	Use it to explain the row without inventing extra interpretation.

What to verify before relying on the data

Methodology explains how a row should be interpreted. Validation shows whether the currently published dataset has enough observations, variation, confidence coverage and transition structure to support downstream use.

Reviewer question	Where to check
Is there enough history for the chain and window I care about?	Validation diagnostics
Do labels vary enough to support segmentation?	Validation diagnostics
Do the fields mean exactly what my integration assumes?	Field Dictionary
Can I inspect a simple joined artifact before integrating?	Analyst Kit
Can I fetch the same structure programmatically?	API Docs