Methodology

Funding rates settle on different intervals (4h or 8h depending on the venue), so each rate is first annualized: rate × (8760 / interval_hours), with the interval inferred from the spacing of a symbol's funding timestamps (default 8h).

We then take the open-interest-weighted average across exchanges: Σ(annualized_funding × oi_usd) / Σ(oi_usd). Venues with no usable open interest are skipped.

The mid-rank empirical percentile of the current aggregated funding within the last 90 days of observations: 100 × (#below + 0.5 × #equal) / N. It is reported only when at least 30 samples are available; otherwise it shows as not available.

Percentage change of total cross-exchange open interest versus the nearest sample around 24 hours / 7 days ago (matched within a ±6h tolerance): (now − then) / then × 100.

The 24h skew between short and long liquidations, in the range −1…1: (short − long) / (short + long). Positive means more shorts were liquidated.

A composite of four normalized components, weighted and then renormalized over whichever components are available (so missing data does not artificially deflate the score):

• Funding extremity (weight 0.35) — distance of the funding percentile from neutral: |pctl − 50| / 50.
• OI momentum (weight 0.25) — 24h OI growth, capped at +50%.
• Liquidation pressure (weight 0.20) — 24h liquidations relative to open interest, capped at 5% of OI.
• Realized volatility (weight 0.20) — hourly log-return volatility, capped at 3% per hour.

Each component is clamped to 0…1; the score is 100 × Σ(weightᵢ × componentᵢ) / Σ(weightᵢ) over the available components.

Funding and open interest are collected every ~5 minutes, klines hourly, and liquidations in real time. Metrics are recomputed on a short cycle; pages revalidate hourly.