Read freshness and evidence quality
without confusing them.

This page answers one question: are the published data files current and usable right now?It does not say anything about what markets are doing or what a subscriber should do.

Each chain shows two separate things that are easy to confuse.

• Health is freshness relative to expected cadence. Bitcoin and Ethereum should update roughly daily. Arbitrum and Base update roughly weekly by design.
• Confidence is evidence quality for the published label. A row can be fresh and still degraded if the evidence surface is weak.

Health is derived at render time from lag versus the chain-specific publication policy. Confidence is read from the latest published meta artifact and measures the evidentiary strength behind the label.

These dimensions are orthogonal. A chain can be delayed but internally coherent, or fresh but epistemically weak. Both are shown to keep operational freshness from masking evidence quality and vice versa.

Different chains publish at different speeds by design. That means lag must always be interpreted relative to policy.

• Bitcoin and Ethereum are expected roughly daily.
• Arbitrum and Base are intentionally published with an expected 7-day delay.

The operational pipeline generally checks for new upstream data twice daily around 09:00 and 21:00 Europe/Oslo.

BTC and ETH use a daily cadence policy: expected 1 day, soft warning above 2 days, hard fail above 4 days. ARB and BASE use a weekly cadence policy: expected 7 days, soft warning above 10 days, hard fail above 15 days.

This is why the same absolute lag value can be normal for one chain and anomalous for another.

Confidence measures how strongly the available data supports the published label.

• Good (≥ 0.70) means strong evidence.
• Caution (0.40–0.69) means moderate evidence.
• Degraded (< 0.40) means weak evidence.

Confidence is shown here as an operational diagnostic, not as a market signal. It indicates how well the evidence surface distinguishes the published label from adjacent labels.

A degraded score means the row may still be published for traceability, but the label should be treated as low-trust.