A source fact is not a conclusion
A price, a quoted spread, or an official economic observation describes something within a defined scope. It does not, by itself, explain what the whole market is doing. DALQEN retains the provider, metric, unit, affected assets, observation time, and collection time before using a fact in an assessment.
Derived intelligence is a separate record. It identifies the evidence and methodology behind a classification. For example, a spread observation from Coinbase Exchange remains specific to that venue. It is not a measurement of liquidity across every exchange.
Freshness depends on the source
Observation time answers when the source measured or reported a fact. Collection time answers when DALQEN received it. Fetching an old observation again does not make it new.
Freshness is evaluated against the rules for the dataset and the assessment window. Official economic releases have different publication schedules from market quotes. A missing weekend release is not treated as though a scheduled market quote had failed.
Historical evaluation also respects when information became available. A fact observed earlier but collected later cannot be used as if the system already knew it at an earlier cutoff.
Quality and coverage stay explicit
The evidence layer distinguishes CURRENT, STALE, PARTIAL, UNAVAILABLE, and LICENSING_RESTRICTED. These states describe whether an input can support the work being done. A reachable source is not automatically current, and a partial response is not complete coverage.
Required windows and compatible inputs matter. If the observations needed for a baseline are missing, DALQEN does not synthesize the missing history. A failed provider check does not replace the last valid value or reset its source timestamp.
Agreement needs independent support
Several measurements from the same underlying dataset do not necessarily provide several independent sources of support. DALQEN tracks independence groups and provider coverage when assessing evidence confidence.
Cross-domain comparisons also need compatible scope, timestamps, and methodology. A market observation and a macro release may add context to each other, but appearing together does not demonstrate causality.
When evidence disagrees, the assessment retains both supporting and conflicting references. Depending on the method and available inputs, the result may be mixed, inconclusive, or insufficient evidence. The system does not force agreement to produce a cleaner narrative.
Confidence describes the evidence
Confidence reflects evidence quality, not return probability or win rate.
The evidence framework uses LOW, MODERATE, and HIGH bands. It considers independent support, freshness, coverage completeness, quality, and agreement. More records alone do not guarantee higher confidence, especially when they repeat the same source or conflict with each other.
These bands are not probabilities that an asset will rise or fall. They do not estimate an expected return or make an assessment a trade instruction.
Insufficient evidence is a valid result
A useful system must be able to stop short of a conclusion. When required coverage or freshness is missing, an insufficient-evidence state records that limitation rather than disguising it as a neutral market reading.
Stable observation identities protect against duplicates. Methodology versions identify how a record was produced, and append-only assessment history preserves earlier results. Later evidence can support a new assessment without silently rewriting the earlier one.
This note describes the existing Evidence Methodology v2 framework. It presents no historical market result, performance study, or directional call. Private assessments remain private, and the Signal Engine remains NOT_ENABLED.
Methodology and sources
Evidence and provenance3 source references
Publication identity and linked records do not establish empirical support. Claim states come from the documented evidence contract; missing references remain missing.
- DALQEN methodology
- Public explanation of evidence handling and confidence.
- DALQEN production source coverage
- Provider scope, collection cadence, and availability limits.
A linked evidence item does not make a claim supported. Missing evidence is not neutral evidence. Timestamp overlap does not establish causality, correlation, or directional market impact.
Risk and limitations
This is a methodology publication, not an empirical market study or investment recommendation. No private evidence records, provider payloads, or research conclusions are reproduced.
Source coverage, availability, licensing, and publication timing can limit an assessment. The linked status page describes current operation and may change after this note is published.
Derived intelligence does not guarantee future outcomes. A methodology description does not establish that every required baseline is mature or that live signal generation is approved.