BtcCrossVenueFundingDiffCarryZScoredDeltaNeutral1H

BTC Hyperliquid-vs-Binance Funding-Differential Carry (Cross-Venue, Delta-Neutral, Z-Scored)

A market-neutral, cross-venue carry strategy that harvests the funding-rate DIFFERENTIAL between the BTC perpetual on Hyperliquid (BTCUSD.HYPERLIQUID) and Binance USD-M (BTCUSDT.BINANCE). The two venues have different participant mixes and funding mechanisms (HL uses a predicted/clamped funding interval; Binance uses premium+interest), so their 8h funding rates routinely diverge. When the differential D = funding_HL - funding_BINANCE stretches to a statistical extreme, the strategy holds a delta-neutral pair — short the high-funding venue, long the low-funding venue, equal notional — collecting the net differential each funding period while carrying ~zero net BTC delta. This fills three under-represented quota buckets at once (cross-venue 5.4%->target 15%, Hyperliquid 7.6%->target 20%, long-short 14%) and deliberately AVOIDS the active failure modes: it is NOT options (dodges the systemic ~0-capital options sizing bug and the structurally-losing naked short-vol family), NOT CASH-spot (both legs are MARGIN perps, dodging the multi-instrument cash-overfill engine defect), NOT a directional funding-fade (dodges the SOL/ETH contrarian falling-knife failure — funding here is a CARRY cash flow, not a sentiment fade), and NOT a single-name HL directional/reversion bet (which is exhausted). Only 2 core parameters to resist overfitting: entry z-threshold and exit z-threshold.

Implements the hypothesis directly. The differential D = funding_HL - funding_BINANCE is read from `hl_funding_rates` and `funding_rates` (the two flat {ns: rate} dual-key dicts), and Binance is wired as the PRIMARY leg precisely because the supplementary loader scopes the top-level `funding_rates` series to the primary venue — with Binance primary, `funding_rates` resolves to Binance and `hl_funding_rates` (resolved by coin, venue-independent) to Hyperliquid, so both legs' funding read correctly; delta-neutrality is unaffected by the labelling. Per the brief the edge is a CARRY cash flow, so D is z-scored (entry_z/exit_z) and the strategy collects the net differential each period rather than fading sentiment — sidestepping the SOL/ETH contrarian falling-knife failure. It deliberately avoids the active failure modes: not options (no ~0-capital sizing bug), not CASH-spot (both legs MARGIN perps, dodging the multi-instrument cash-overfill defect), and a hedged two-leg book rather than a single-name directional perp (dodging the over-exposure engine bug). Only 2 core tunables resist overfitting; fixed equal-notional sizing with leverage=1.0 bounds exposure. calculate_signal returns the raw z (natural units, unbounded) thresholded in should_enter, and returns 0.0/flat with NO price-only fallback when funding is missing, per the no-silent-fallback rule. Fills the cross-venue, Hyperliquid, and long-short quota buckets at once.

Not worth optimizing — no realized edge after two-venue costs, and the only positive number is a suspected artifact. Realized record is a consistent bleed: 8 straight negative months (-11% to -58%), profit_factor 0.99, Sharpe -0.49, Sortino -3.27, negative expectancy, win_rate 47.9% with avg_win~=avg_loss. The headline +81% total_return is entirely a single +315% final-bar day with end_unrealized_pct=1.015 (~100% of return is un-exited open-position MTM); a +315% day on a ~zero-net-delta pair is implausible and points to a leg-netting/cross-venue alignment artifact (symptom stated, root cause not independently verified — escalate for harness review rather than guess). The BTC major-venue funding differential is real but too small to clear ~0.10% (Binance) + ~0.09% (HL) round-trip per pair, so the carry loses across every regime in the sample. impact_cost_pct 122% and capacity_usd $66k confirm it works only at toy scale. This is NOT a funding-uncreditable false negative (funding is now credited) and NOT a parameter problem entry_z/exit_z tuning can fix — it is structural for BTC. Possible Research-Lead reframe (revise_hypothesis territory, not iterate): retry the same cross-venue funding-differential mechanism on a higher-funding-skew instrument where the differential exceeds two-venue costs.

Delta-neutral cross-venue funding-differential carry on BTC perpetuals: long/short equal notional on BTCUSD.HYPERLIQUID vs BTCUSDT.BINANCE (both MARGIN perps). Each bar it computes D = funding_HL - funding_Binance from the supplementary funding series (nearest-timestamp lookup, 8h tolerance for Binance / hourly for HL), standardizes D into a rolling z-score over lookback_bars, and trades the statistical extreme: when z > entry_z (HL funds richer) it goes LONG Binance / SHORT HL; when z < -entry_z (HL funds cheaper) it goes SHORT Binance / LONG HL; it closes the pair when |z| < exit_z (convergence) or after max_hold_hours. Both BTC legs are equal-and-opposite, so net BTC delta is ~zero and the book harvests only the funding differential. Entry is atomic — both legs validated before either order — to avoid a naked primary leg, and exits close both legs in lockstep. Sizing is fixed equal-notional per leg (equity * position_frac), never scaled by signal or leverage (leverage=1.0). Only 2 core edge parameters: entry_z and exit_z.

Genuinely market-neutral construction (beta -0.0045, benchmark_meaningful=false correctly flags buy-hold as the wrong baseline); both legs MARGIN perps so it dodges the CASH-overfill engine defect

Atomic two-leg entry and funding-read plumbing work: 72/72 entries submitted, balanced 72 long / 72 short, funding credited on both venues

Only 2 core tunables (entry_z/exit_z) — low overfitting surface

Headline +81% total_return is an artifact: 8 consecutive negative months overturned only by a +315% final-bar day; end_unrealized_pct=1.015 means ~100% of the return is unrealized open-position MTM at backtest end, never realized

+315% single day on a delta-neutral pair is implausible — suspected leg-netting/cross-venue alignment artifact (mechanism not independently verified)

No realized edge after costs: profit_factor 0.99, Sharpe -0.49, Sortino -3.27, negative expectancy, commission 10.2% of gross — the BTC HL-vs-Binance differential is smaller than two-venue round-trip fees

impact_cost_pct 122% and capacity_usd only $66k — edge (if any) exists only at toy scale; position_frac 0.40 is oversized

8/9 months negative

majority positive

BtcCrossVenueFundingDiffCarryZScoredDeltaNeutral1H tried to harvest the funding-rate differential between BTC perps on Hyperliquid and Binance as a delta-neutral, low-parameter carry, deliberately dodging the options, CASH-overfill, and directional-funding-fade failure modes. The plumbing worked — atomic two-leg entries, balanced 72 long / 72 short, funding credited on both venues, genuinely market-neutral (beta -0.0045) — but the realized edge was negative: 8 straight losing months, profit factor 0.99, and fees at 10.2% of gross, because the BTC differential is smaller than the two-venue round-trip cost. The flashy +81% headline was dismissed as a suspected harness artifact (a single +315% final-bar day, ~100% unrealized) and escalated for review rather than trusted. The reviewer abandoned it as structurally unprofitable on BTC, suggesting the research lead reframe the same mechanism onto a higher-funding-skew instrument where the differential could actually clear costs.

The BTC major-venue funding differential is too small to clear two-venue round-trip fees, so cross-venue funding carry is structurally unprofitable on BTC and not fixable by z-threshold tuning — but the mechanism and clean delta-neutral two-leg construction are sound, so a revise-hypothesis retry on a higher-funding-skew instrument (where the differential exceeds costs) is the path, not iteration.

The analyst abandoned it at the backtest-review gate, before optimization, because there is no realized edge after two-venue costs — the BTC funding differential is real but structurally smaller than the ~0.10% + ~0.09% round-trip, so it loses across every regime — and the only positive number is a suspected leg-netting/cross-venue alignment artifact escalated for harness review rather than trusted.

A market-neutral, delta-neutral cross-venue carry strategy that harvests the BTC funding-rate differential between Hyperliquid and Binance USD-M perps — z-scoring D = funding_HL − funding_BINANCE and holding an equal-notional opposite-side pair (short the richer-funding venue, long the cheaper) at statistical extremes — with only 2 core tunables (entry_z, exit_z) to resist overfitting.

Its realized record was a consistent bleed: Sharpe -0.49, Sortino -3.27, profit factor 0.99, avg_trade_return -$2.82 over 144 trades (72 balanced pairs), 8 straight negative months, fees 10.2% of gross; the +81% headline total return was an artifact — ~100% unrealized open-position MTM (end_unrealized_pct 1.015) driven by a single implausible +315% final-bar day on a zero-net-delta pair — and impact_cost_pct 122% with capacity_usd $66k confirm toy scale only.