Strategies

Hyperliquid LTC–BCH Low-Participation Relative Divergence Reversal

Outcome: Abandoned

HyperliquidLtcBchLowParticipationReversal

Outcome Summary

This hypothesis tested whether a three-day LTC-versus-BCH price divergence on low volume reverses. It traded synchronized 4-hour Hyperliquid perpetual bars, with catalog coverage from 2024-01-25 to 2026-10-02. The developer coded HyperliquidLtcBchLowParticipationReversal, a long-laggard/short-leader strategy with equal dollar notionals, but over 3 iterations it never got past verification. Each time the Layer 1.5 consistency check flagged a hypothesis mismatch because the optimization_plan fixes leverage at 1.0. The run was abandoned before any backtest, optimization, analyst or risk review.

Strategy report

Hypothesis

Test whether substantial three-day LTC-versus-BCH price divergence reverses when neither asset attracts above-normal trading volume. Use synchronized 4-HOUR perpetual bars, buying the relative loser and shorting the winner with equal dollar notionals. HyperliquidCollector supplies OHLCV for both exact instruments. Catalog inspection verified 5,883 matching timestamps covering the complete shared four-hour grid from 2024-01-25 16:00 UTC through 2026-10-02 00:00 UTC, with timestamp markers present. Library inspiration: https://www.fmz.com/strategy/426047, whose price-volume interpretation motivates distinguishing low-participation moves; this proposal does not copy its RSI/Bollinger implementation. Research background: [Dynamic Volume-Return Relation](https://www.nber.org/papers/w8312) distinguishes risk-sharing reversal from information-driven continuation. Applying that distinction to these crypto perpetuals is an unvalidated hypothesis.

Hyperliquid accounts for 7.3% of experiments against a 20% target, while long-only strategies account for 83.5% against a maximum target of 55%. This two-instrument long-short experiment addresses both imbalances and avoids additional BTC testing. Corpus concentration was empty, so instrument-specific usage cannot be asserted. Unlike the failed AAVE–UNI stress continuation and sector lead–lag proposals, entry requires low participation and trades reversal of relative divergence without a joint-selloff condition. Options quotas are deferred because the supplied outcomes show structural options-cost failures. Funding carry is excluded because the consulted specification says funding payments are not credited by the backtest engine.

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

Implementation

Trades equal-dollar LTC/BCH perpetual legs against three-day relative divergence when both assets have low participation. Exits on half recovery, adverse spread movement, holding expiry, or an 8% individual-leg loss. Spread and emergency stops are checked at synchronized four-hour closes; intrabar losses and execution gaps can exceed those thresholds. Residual-leg recovery retries closes on subsequent bars until both instruments are flat and outstanding orders are resolved.

Preserves the previous continuous signal, synchronized observations, rolling volume calculations, sizing, fixed leverage and research parameter permissions. Corrects emergency-stop units by converting the 0.08 fraction to 8 percentage points: a helper result of -1.0 does not trigger, whereas -8.0 does. Adds a persisted scalar closing state, independent secondary-bar reconciliation, and position-close/rejected/denied-order callbacks that flag recovery. Closing state remains active through rejected, skipped or pending closes, blocks new entries, cancels unfinished entries, and clears only after both legs and outstanding orders are resolved. Retries occur on bars to avoid recursive rejection-driven submissions. The reported negative economics require reassessment after this material stop correction; they do not yet establish that the intended mechanism lacks an edge. No verification or final-exam data reads were performed. The requested outbox file could not be written because this session permits read-only filesystem access; the complete bound JSON is returned here.

Verification Results

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

Iteration History

Verification failed (Layer 4 — QA review) [class=code_defect]: - [edge_concern] The sandbox reports 356 closed positions, profit factor 0.426 and average net return -0.478% per position. Reassess economics after correcting the emergency-stop units; these results do not determine the correctness verdict. - [critical] get_position_pnl_pct() returns percentage points: a 1% loss returns -1.0. Comparing it against emergency_leg_stop_pct=0.08 therefore exits at a 0.08% leg loss rather than the intended 8%. This makes the emergency stop 100 times tighter and prematurely closes the pair before the spread-recovery mechanism can operate. (line 196) — fix: Multiply emergency_leg_stop_pct by 100 before both comparisons, or divide both helper results by 100. Verify that a 1% leg loss does not trigger and an 8% loss does trigger. - [critical] Orphan-leg recovery clears _active immediately after submitting a close, before confirming that the residual position is flat. If BCH remains open because its close is rejected or was skipped while an order was working, subsequent recovery is disabled. should_enter() blocks on the BCH position, and the base dispatch never calls should_exit() without an LTC position, leaving the unhedged BCH leg unmanaged. (line 118) — fix: Maintain an explicit closing/recovery state until both instruments are flat and outstanding entries are resolved. Reconcile and retry residual closes independently of whether the primary position exists, preferably through order and position callbacks. - [warning] Spread and emergency stops are checked only at synchronized four-hour closes, so intrabar losses can exceed their thresholds. (line 192) — fix: Document the stops as close-based and assess intrabar exposure, or provide compatible protective orders and more frequent paired risk monitoring.

Outcome Summary

Make the leverage setting in the optimization_plan agree with the hypothesis's declared risk profile before submitting, so the Layer 1.5 check passes and the strategy can reach a backtest.

It was abandoned in a verification loop after 3 iterations. Layer 1.5 (the check that the config matches the hypothesis) kept failing with class hypothesis_mismatch because the optimization_plan fixes leverage at 1.0.

An equal-dollar pairs trade on Hyperliquid LTC and BCH perpetuals using 4-hour bars: after a large three-day price divergence that happened without above-normal volume in either coin, buy the laggard and short the leader, expecting the gap to close.

No performance numbers exist. The backtest report is empty, there is no optimization report, and the strategy never reached backtesting.

Backtest and paper results are hypothetical. Trading involves risk of loss.