HyperBasis Methodology

Every number HyperBasis publishes has a precise definition. This page documents the formulas and the safeguards around them.

1. Deviation (Hyperliquid vs TradFi)

Deviation measures the percentage dislocation of the Hyperliquid mark price against the TradFi reference:

deviation% = ( hl_effective − tradfi_price ) / tradfi_price × 100

where hl_effective is the Hyperliquid mark price divided by the ADR ratio when the perp's quote scale differs from the underlying (for example, a perp quoted ~90x the ETF price). A positive deviation means Hyperliquid trades at a premium; negative means a discount.

2. Alignment engine

Hyperliquid prices are chain-time; TradFi prices are exchange-time. The alignment engine pairs each Hyperliquid minute with the correct TradFi bar using one of four methods:

• exact — a TradFi bar exists at the identical timestamp.
• ffill — the nearest prior TradFi bar within 90 seconds is used (forward-fill).
• vwap_interp — the Hyperliquid timestamp falls between two TradFi bars; the price is volume-weighted interpolated.
• market_closed — no TradFi bar within 90 seconds (weekend/halt). These bars are deliberately NOT emitted, because writing a deviation there would fabricate a signal.

3. Staleness guard

Every live TradFi quote carries the market timestamp of the quote itself. If the quote is older than 5 minutes, the resulting deviation row is flagged TRADFI_STALE. The terminal displays the flag and the timestamp; stale rows never drive real-time signals. This closes the classic "today vs yesterday" trap where an off-hours close looks fresh.

4. Funding APR

Hyperliquid publishes a funding rate per instrument per hour. HyperBasis annualizes it:

funding_apr% = funding_rate × hours_per_year

and records snapshots every 10 minutes into a history that powers percentile and z-score analysis of funding regimes.

5. Capacity score

The capacity score estimates how much size an instrument can absorb before the depth-1% price impact becomes material. It combines L2 orderbook depth within 1% of mid, open interest, and historical volume into a normalized score. Higher scores mean deeper, more absorbent books.

6. Z-scores and percentiles

For each asset, HyperBasis maintains rolling statistics (mean, standard deviation, percentiles) of deviation and funding. The current value is expressed as a z-score — how many standard deviations from the 90-day mean — and as a percentile of the historical distribution. Assets with fewer than the minimum history points are labeled with reduced confidence rather than being silently included.

7. Anomaly marking

Deviation rows beyond the historical anomaly threshold (extreme oracle errors or liquidity flashes) are flagged anomaly=1 and excluded from statistics, with an audit trail. Users can query real-only history to exclude flagged bars.

8. Daily z-score (transitional layer)

While minute-level history accumulates toward the full 90-day standard, HyperBasis computes a close-price-based daily z-score per asset as a transitional statistical layer. It is always labeled as daily granularity and never presented as minute statistics.