METHODOLOGY

PerpMetric Research Methodology

PerpMetric uses market-relative baselines, explicit data-quality states and completed forward windows. This page describes the interpretation model behind the interface.

Published and reviewed 2026-09-11

1. Normalize the market identity

Core and HIP-3 instruments are assigned canonical market identifiers. This matters because the same display symbol can exist on different builder-deployed DEXs with separate order books, oracle definitions and market settings.

2. Build recent reference distributions

Price returns, open-interest changes, funding and mark-oracle differences are compared with recent observations for the same market. Percentile regions and robust standardized distances describe how unusual the current value is. They do not assign directional certainty.

3. Preserve the condition

An event records the metric, direction, severity path, reference window and market set. Keeping these dimensions explicit prevents unlike situations from being silently combined.

4. Complete forward horizons

Returns are written only when the required future timestamp can be anchored to collected market data. Research summaries use completed observations and report missing anchors or targets as data exceptions rather than filling them with invented values.

5. Summarize distributions

The interface emphasizes median, interquartile range, positive share, sample size and coverage. These are descriptive statistics. They can support hypotheses and strategy research, but a production strategy still needs out-of-sample validation, realistic fees, slippage, liquidity limits and risk controls.

Inspect the resulting research

Primary references

Audience measurement

We measure visits to public pages, choices of instruments and research, page sections, broad device and browser categories, and estimated active use while a page is visible, focused and recently used. Background price updates are counted separately from instrument choices.

These measurements use our own infrastructure. We do not use analytics cookies, browser storage, cross-site tracking or persistent visitor identifiers. We do not measure individual returning visitors or personal browsing journeys. Visitor estimates use network addresses temporarily during daily calculation; the retained totals contain no IP addresses or complete browser signatures.

Server access logs can contain network addresses and user-agent headers and rotate under the existing log policy (up to 14 retained files and 14 days). Daily anonymous totals are retained without automatic deletion. Browser measurements can be blocked or incomplete. We do not collect free-text searches or location data for these statistics.