HyperliquidCrossSectionalMomentumDollarNeutralBasket

Hyperliquid Cross-Sectional Momentum Dollar-Neutral Basket (Relative-Strength Rank, Weekly Rebalance, Long-Short)

A dollar-neutral, multi-instrument CROSS-SECTIONAL (relative-strength) momentum basket across the ~15 most liquid Hyperliquid perpetuals. Unlike time-series/absolute momentum (which holds each asset long or short on its OWN trend — the mechanism already promoted to paper this session), cross-sectional momentum ranks assets AGAINST EACH OTHER and is always long the strongest relative performers and short the weakest, regardless of the broad market's absolute direction. This makes it a genuinely complementary alpha sleeve, not a duplicate: it harvests dispersion (winners keep outperforming losers) rather than directional trend, so it is structurally market-neutral and low-correlation to both the portfolio's dominant long-only book and the promoted absolute-momentum basket. It deliberately avoids every refuted mechanism this session — no funding carry, no spot-perp/cross-venue basis, no options (now flagged unbacktestable). Both long and short legs execute on the SAME venue (Hyperliquid) at the same daily bar timestamps, so there is no cross-exchange fill-timing artifact. Pure OHLCV, few parameters, weekly rebalance for low turnover. Fills the under-represented Hyperliquid venue, multi-instrument scope, and long-short direction buckets at once.

Iteration 2 fix: the only change from previous_code is converting `_bar_ts` from a `@staticmethod` to a regular instance method `def _bar_ts(self, bar)`. The Layer-2 synthetic proxy rebinds strategy methods as instance methods, which broke the staticmethod call (`self._bar_ts(bar)` passed both self and bar -> 'takes 1 positional argument but 2 were given'). As an instance method it now receives (self, bar) correctly. All signal, ranking, dollar-neutral sizing, and rebalance logic that passed Layer 1 is unchanged. Implements the hypothesis: cross-sectional relative-strength ranking (long strongest, short weakest) rather than absolute trend, both legs on Hyperliquid at aligned daily timestamps, pure OHLCV, weekly rebalance, no funding/basis/options mechanisms.

Not worth optimizing despite a healthy 431-trade sample and a correctly-implemented, genuinely market-neutral construction. Two structural problems that the 3 tunables (lookback, n_legs, rebalance_days) cannot fix: (1) CAPACITY/IMPACT WALL — impact_cost_pct is 40% with capacity_usd only $623k, because the cross-sectional mechanism requires a broad universe but the HL alt cohort (ATOM/NEAR/APT/ARB/OP/INJ/SUI/BCH) is too illiquid; ~40% of gross PnL is eaten by market impact and the edge exists only at toy scale, a hard promotion blocker. Restricting to liquid HL majors to cut impact would leave too few names to rank (k=4 needs >=8), defeating the mechanism. (2) THIN, DECAYING EDGE — base profit_factor 1.03 and Sharpe 0.18 (sharpe_ci_low -0.84, straddling 0), with annual returns falling from +22.7% in 2023 (the HL-launch dispersion window) to -2.9% (2024) and -8.5% (2025); the dispersion premium faded as the venue matured. A best-of-225 optimization on this weak a base would inflate in-sample Sharpe but fail the downstream deflated-Sharpe/holdout gates, and the illiquidity/capacity ceiling would block deployment regardless. The session already promoted an absolute-momentum HL basket filling the HL/momentum bucket, so this adds no deployable complementary sleeve. (Note: abandon is for the capacity wall + decayed dispersion edge on an illiquid cohort — the mechanism itself is sound and could be revisited on a more liquid multi-venue universe.)

Dollar-neutral cross-sectional (relative-strength) momentum basket across ~15 liquid Hyperliquid perpetuals on daily bars. Every weekly rebalance it ranks all valid assets by their lookback return and goes long the top k and short the bottom k in equal dollar notional per leg, so long notional == short notional (structurally market-neutral). Gross exposure hard-capped at gross_cap (1.0x = no leverage). Same-side books resize only on a material delta; side flips close then reopen.

Verification failed (Layer 2 — synthetic scenarios): Parameters used: ['assets', 'n_legs', 'lookback', 'gross_cap', 'skip_days', 'resize_frac', 'min_notional', 'qty_precision', 'rebalance_days', 'default_qty_precision'] Check that __init__ sets all attributes from self.parameters.get(). - steady_uptrend: TypeError: HyperliquidCrossSectionalMomentumDollarNeutralBasket._bar_ts() takes 1 positional argument but 2 were given (bar timestamp: 1735691460000) - steady_downtrend: TypeError: HyperliquidCrossSectionalMomentumDollarNeutralBasket._bar_ts() takes 1 positional argument but 2 were given (bar timestamp: 1735691460000) - flat_ranging: TypeError: HyperliquidCrossSectionalMomentumDollarNeutralBasket._bar_ts() takes 1 positional argument but 2 were given (bar timestamp: 1735691460000) - volatility_spike: TypeError: HyperliquidCrossSectionalMomentumDollarNeutralBasket._bar_ts() takes 1 positional argument but 2 were given (bar timestamp: 1735691460000) - zero_volume: TypeError: HyperliquidCrossSectionalMomentumDollarNeutralBasket._bar_ts() takes 1 positional argument but 2 were given (bar timestamp: 1735691460000) - price_gap: TypeError: HyperliquidCrossSectionalMomentumDollarNeutralBasket._bar_ts() takes 1 positional argument but 2 were given (bar timestamp: 1735691460000)

Clean, differentiated construction: genuinely dollar-neutral (beta -0.002, benchmark_correlation -0.009), balanced long/short (195/236 trades), same-venue same-timestamp execution (no cross-exchange fill artifact), positive alpha — the mechanism is implemented correctly and complements the promoted absolute-momentum sibling in principle

Healthy trade count (431) over a multi-year daily sample — sample size itself would support optimization

Structurally un-deployable: impact_cost_pct 40% and capacity_usd only $623k — the illiquid HL alt cohort means market impact consumes ~40% of gross PnL and the edge exists only at toy scale (hard capacity blocker, not fixable by the 3 tunables)

Thin, decaying edge: profit_factor 1.03, Sharpe 0.18 with sharpe_ci_low -0.84 (straddles 0), and annual returns collapse from +22.7% (2023) to -2.9% (2024) and -8.5% (2025) — the dispersion premium faded as Hyperliquid matured

High drawdown (31.4%) and outlier-carried tail (kurtosis 72, skew 4.2, tail_ratio 0.36) for a supposedly market-neutral book

Best-of-225 optimization on this weak a base would fail the downstream deflated-Sharpe/holdout gates — a 2-hour sweep is very unlikely to yield a promotable, deployable config

This strategy aimed to add a genuinely complementary, market-neutral sleeve by ranking the most-liquid Hyperliquid perps against each other and trading winners-versus-losers on a single venue. The mechanism was implemented cleanly — dollar-neutral, balanced legs, positive alpha over 431 trades — but the economics did not hold: a 0.176 Sharpe and 1.03 profit factor sat atop a hard capacity wall where 40% of gross PnL was lost to impact at only ~$623K capacity, and the edge decayed from +22.7% in 2023 to -8.5% by 2025 as the venue matured. The analyst abandoned it at the pre-optimization gate after 2 iterations, judging that no parameter sweep could fix an illiquidity ceiling and a fading premium, and that the already-promoted absolute-momentum HL basket left it adding no deployable value. The mechanism itself was deemed sound and potentially revisitable on a more liquid, multi-venue universe.

A cross-sectional dispersion mechanism needs a broad universe to rank, but on an illiquid venue that breadth forces trading names whose market impact eats ~40% of gross PnL — restricting to liquid majors leaves too few names to rank — so the construction can be correct and market-neutral yet remain undeployable, especially once the dispersion premium decays as the venue matures.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate: two structural problems the three tunables cannot fix — a capacity/impact wall (40% impact cost, ~$623K capacity on the illiquid HL alt cohort) and a thin, decaying dispersion edge — and the session had already promoted an absolute-momentum HL basket filling the same bucket. Optimization and all later stages were never reached.

A dollar-neutral cross-sectional (relative-strength) momentum basket across the ~15 most-liquid Hyperliquid perpetuals that each weekly rebalance ranks assets against each other and goes long the strongest k and short the weakest k in equal dollar notional, harvesting dispersion rather than directional trend as a low-correlation complement to the promoted absolute-momentum sleeve.

Over ~3 years (2020-08-19 to 2026-05-20) it took 431 trades (195 long / 236 short), was genuinely dollar-neutral (beta -0.002, benchmark correlation -0.009), and returned +16.6% total (CAGR 4.45%) — but with only a 0.176 Sharpe (CI -0.84 to 1.01), profit factor 1.03, and a 31.4% max drawdown. Critically, market impact consumed ~40% of gross PnL with capacity only ~$623K, and annual returns decayed from +22.7% (2023) to -2.9% (2024) to -8.5% (2025).