← Research

Why an edge should exist / hypothesis

A public ID removes the core barrier of ride-the-whale and single-venue absorption: ordinary tape carries no participant ID. Testable hypothesis: transparency yields a tradeable edge in directional strategies.

Construction

Four angles. (1) Follow-the-whale v2: MM filter, position-level equal-weight with a cap, de-beta vs BTC. (2) Reverse-engineer the style of profitable non-MMs. (3) Per-address accumulation detection with a causal pre-signal gate. (4) Liquidations and forced flow as a spike fade.

Test data

SQD Portal (a free HTTP source of all HL fill history with addresses): leaderboard of 44,923 addresses, 44 thin coins × 21 days = 41.2M fills, a cohort of 500 ordinary accounts $5k–150k.

Results

Follow v2: H1→H2 persistence +42 bps, but 95% CI [−110, +191], the entire gain on a single address. Ordinary directional retail: 52% net-profitable, median +$108 (about zero), cohort −$717k over 180 days; ~19% of activity is farm/hedge. Accumulation: reverse causation — the accumulator chases an already-past +28 bps, forward is −4.8. Liquidations: the fade is real (+13…+19 bps with a stop), but it is the existing spike edge.

Validation

follow, reverse and accum are null both on history and on persistence; all tested cheaply on history, with no weeks of forward collection and no AWS.

Root cause

The same transparency makes any signal crowded (adversarial) and does not solve the core issue: we see only the HL leg, not the intent. Leaderboard winners are hot-month variance plus pure holds where the decision was made before the window and is unobservable.

Verdict

Dead for directional strategies. A public ID gives no tradeable edge; the only real reversion is a generic spike fade already traded on Bybit — a capacity-extension lead, not new alpha.

Source

archive/HYPERLIQUID_RESEARCH.md · PROGRESS 2026-09-04