Accuracy, not uptime
Most vendor evaluations measure availability. This measures whether the answer was correct, scored against chain ground truth. The two axes are complementary and independent: a provider can be perfectly healthy and consistently wrong. Operational freshness lives inget_data_source_health; accuracy lives here.
Five query classes
Providers are scored on question types where a chain-derived answer exists to check against — wallet labelling, volume, first-seen timing, ordering, and staleness.explain_query_class returns the definition for any class.
A provider is never scored against itself
A vendor evaluation that quietly ranked all six would be more satisfying and completely wrong.Distinctions the verdicts preserve
Five verdicts, not two, because the ways an answer can fail to be right are not interchangeable:Reading the board
null means unmeasured
Never zero. A provider with
null on a class was not tested on it.Every rate has a denominator
Check
scored before quoting hit_rate. A perfect rate over one sample is not a record.30d, 90d, and all.
Three data modes
A board can be demonstration data, measured-with-zero-rows, or measured-with-results. The middle state is normal before the first credentialed run — the methodology and query definitions are real, there are simply no measurements yet. See the endpoint reference.Access
Full specification:
docs/methodology/provider-reliability-v1.md.Scienter publishes impersonal market signals of general and regular circulation. Nothing here is investment, legal, or tax advice, or a recommendation to buy, sell, or hold any asset. Scienter is not a registered investment adviser or broker-dealer. Trading digital assets can result in the total loss of your funds — see Disclaimers for the full text.