Hyperliquid Funding-Crowding Stall Breaks: HBAR, FIL and RENDER Long/Short

Hypothesis

Test whether persistently expensive one-sided perpetual positioning that cannot extend price unwinds after price breaks against the funding-paying side. Trade three independent Hyperliquid perpetuals using 12-HOUR OHLCV bars and settled hourly funding. This is a directional positioning hypothesis, not funding carry. Market orders execute at the completed signal-bar close, explicitly using the registry's continuous-market equivalent for next-bar-open execution. Protective stop-market orders trigger intrabar through platform matching. Use each instrument's actual overlapping bar/funding history. Predict 10–25 qualifying episodes per instrument-year; this is unmeasured. No preliminary performance test was run.

Long-only exposure is 77% against a 55% ceiling, making direction the largest quota gap. This symmetric panel also adds underrepresented Hyperliquid and multi-instrument exposure. The concentration map is empty; the available snapshot shows only 3 HBAR, 4 FIL and 1 RENDER primary-symbol runs. Prior-art searches found no funding-stall/absorption hypothesis on these instruments. Unlike abandoned funding-sign-transition strategies, funding must remain one-sided at entry and price must stall before breaking against the paying side. Unlike the NEAR spot-confirmed funding trap, this design requires neither spot confirmation nor an already-developed opposing multi-day move.

Reject advancing this measured three-instrument baseline: average net trade return is -0.3438% versus the +0.15% Hyperliquid review reference, with losses in HBAR and RENDER. Aggregate realized PnL is positive because larger FIL winners outweigh those losses; aggregate negative expectancy is not established. The weak typical-trade economics and concentrated gains do not justify optimizing this baseline without a concrete improvement hypothesis. This verdict does not establish that funding-crowding stall breaks generally lack an edge.

Implementation

Three independent 12-hour Hyperliquid funding-crowding stall-break legs. Persistent positive funding followed by a stalled range and downside close break enters short; the symmetric negative-funding upside break enters long. Market orders fill at signal-bar close, the registry continuous-market equivalent of next-bar-open execution. Each entry receives a 6% intrabar protective stop and exits after hold_bars own-instrument completed bars.

Operationalizes stall as the prior stall_bars high-low range <= stall_atr_mult times the prior 14-bar simple-average true range; breakout is a close beyond that prior range. Funding daily cost is 24 times the mean settled hourly rate over funding_days, with 90% hourly-slot coverage, 75% paying-side sign share, latest rate still on that side, and at most two hours old. No price-only funding fallback or cross-leg signal contamination. Uses each leg only where its own bar/funding history overlaps. Sizes to 0.5% equity risk at the 6% stop, capped by 10% leg exposure, 30% gross exposure and 0.1% signal-bar volume. Holding counters are per-leg scalar attributes persisted across restarts. Baselines use the centers of the registered tunable domains; all fixed research values are preserved. The episode-rate prediction remains unmeasured.

Verification Results

Sandbox average net trade return is -0.3438% while total return is +0.8989%. Reconcile entry-notional weighting, leg-level PnL, commissions and funding in the full-history ledger before drawing economic conclusions. The aggregate-return deployment heuristic is not interpretable with opposing metric signs; reported average position size is 2.0322% of equity.

Backtest Review

Sharpe

0.11

Total return

0.90%

Max drawdown

4.93%

Trades

24

Win rate

33.3%

Profit factor

1.54

Closed-trade PnL reconciles to +$898.88, including +$66.23 funding. Entry-notional-weighted net return is +1.846%, explaining the positive portfolio return despite the negative arithmetic trade average. No accounting defect is established.

Code and trade patterns are consistent with funding-conditioned opposing range breaks, variable sizing, five-day timed exits and approximately 6% stop exits.

Average net trade return is -0.3438%, below the +0.15% Hyperliquid review reference. QA’s opposing-sign concern is reconciled, but weak typical-trade economics remain.

RENDER loses $838.24 and HBAR loses $249.52; FIL gains $1,986.65. One FIL winner contributes $1,494.01; excluding it leaves -$595.13. This indicates concentration, not an invalid trade.

The 24 trades represent approximately 3.7 per instrument-year versus the unmeasured prediction of 10–25. Scarcity alone is not the rejection reason.

No demonstrated implementation defect or evidence-backed tuning direction supports iteration. The registered parameter space remains untested.

The supplied benchmark is not meaningful for this book and is not used as a rejection criterion.

Outcome Summary

A future iteration needs a concrete improvement hypothesis addressing weak typical-trade economics and concentrated gains before optimization is justified.

Backtest review abandoned the baseline because average net trade return fell below the +0.15% reference, HBAR and RENDER lost money, and gains were concentrated in FIL without a concrete improvement hypothesis. Optimization, post-optimization analysis and risk review were not reached.

It traded long and short Hyperliquid HBAR, FIL and RENDER perpetuals when 12-hour price ranges broke against persistently funding-paying positioning.

Over 790 days, the backtest produced 24 closed trades, a 0.8989% total return, Sharpe of 0.1086 and 4.9303% maximum drawdown. Net realized PnL was +$898.88, but average net trade return was -0.3438%; excluding the largest FIL winner left -$595.13.