Hyperliquid DeFi Liquidity Recovery versus Deterioration

Hypothesis

Test whether improving trading liquidity predicts multi-day relative appreciation across COMP, CRV, DYDX and SNX perpetuals. Use synchronized 4-HOUR OHLCV collected by HyperliquidCollector. Exact catalog inspection verified 5,897 common timestamps from 2024-01-22T12:00:00Z through 2026-10-02T00:00:00Z, versus 5,902 expected grid points. Trade one equal-dollar long-short pair at a time. The mechanism is a change in within-token return-per-dollar-volume, rather than price momentum, spread mean reversion or funding carry. Amihud's liquidity measure supplies the measurement idea; its equity evidence does not establish this crypto hypothesis. Source: https://doi.org/10.1016/S1386-4181(01)00024-6. Implement with bounded rolling buffers and FactoryStrategy multi-instrument hooks.

Long-only exposure is the largest policy imbalance: 83.4% against a maximum target of 55%. This proposal adds opposing positions and multiple instruments on Hyperliquid, whose 7.4% share is below its 20% target. Corpus concentration was supplied as an empty object, so individual-symbol research shares cannot be established. Avoid options given the structural failures, and avoid pure funding carry because the reference document says backtests do not credit funding cash flows. Same-venue execution keeps the experiment simpler than the failed cross-venue convergence trades. This tests liquidity repricing without requiring a preceding price shock, a calendar trigger or supplementary confirmation.

verification_loop: sandbox_timeout: Verification failed (Layer 3 — sandbox backtest): smoke test exceeded the 300s wall-clock limit on a ~20000-bar window (2400 days of COMPUSD.HYPERLIQUID-4-HOUR-LAST-EXTERNAL). This almost always means per-bar work that scales with history — e.g. rescanning the full funding/supplementary series, or rebuilding a list and calling min()/sorted() inside calculate_signal()/on_bar() on every bar. Precompute sorted timestamp arrays ONCE in __init__ and use bisect, or cache lookups keyed by timestamp, so per-bar cost is O(log n) not O(n).

Implementation

Trades one equal-dollar Hyperliquid pair across COMP, CRV, DYDX and SNX: long the strongest liquidity recovery and short the strongest deterioration, measured by the log ratio of preceding-baseline to recent Amihud means. Uses turnover eligibility, volatility-based pair sizing, a spread-loss stop and a multi-day holding limit.

Addresses the reported Layer 1.5 mismatch by explicitly setting leverage to the fixed 1.0 in both config and parameters and preserving the exact research optimization plan. Previous signal, alignment, bounded rolling calculations, entries and exits are preserved; the source change only documents the fixed leverage beside its sizing reference. No independent verification was run and no final-exam data was read. The required outbox file could not be written because this session has read-only filesystem permissions; the complete bound JSON is returned here.

Verification Results

sandbox_timeout: Verification failed (Layer 3 — sandbox backtest): smoke test exceeded the 300s wall-clock limit on a ~20000-bar window (2400 days of COMPUSD.HYPERLIQUID-4-HOUR-LAST-EXTERNAL). This almost always means per-bar work that scales with history — e.g. rescanning the full funding/supplementary series, or rebuilding a list and calling min()/sorted() inside calculate_signal()/on_bar() on every bar. Precompute sorted timestamp arrays ONCE in __init__ and use bisect, or cache lookups keyed by timestamp, so per-bar cost is O(log n) not O(n).

Iteration History

Verification failed (Layer 1.5 — hypothesis/config consistency) [class=hypothesis_mismatch]: - optimization_plan fixes leverage at 1.0

Outcome Summary

In a multi-instrument strategy, per-bar cost must stay flat as history grows. That means fixed-size incremental state and no repeated rebuilding or re-sorting of bar histories or timestamp sets (such as the dict/set replay in _seed). Measure the runtime on a full-size window before resubmitting.

After 3 iterations the run was abandoned in a verification loop: the Layer 3 sandbox backtest went over the 300s wall-clock limit on a ~20,000-bar (2,400-day) COMPUSD.HYPERLIQUID 4-HOUR window. Backtest review, optimization, analyst and risk review never ran.

An equal-dollar long-short pair across Hyperliquid COMP, CRV, DYDX and SNX 4-hour perpetuals, betting that tokens whose Amihud-style illiquidity (return per dollar volume) is improving relative to their own baseline would outperform tokens whose liquidity is deteriorating over multi-day horizons.

No performance figures exist: the backtest report is empty and there is no optimization report, so there are no return, Sharpe or trade-count numbers.