Cross-Venue Spot-Anchored Funding Carry: Long BTC Binance Spot vs Short BTC Hyperliquid Perp, Delta-Neutral (Harvest HL Hourly Funding Premium, Avoids the Broken USD-M Perp Leg, Low-Parameter)

A DELTA-NEUTRAL, CROSS-VENUE cash-and-carry strategy that holds a REAL spot long on BINANCE_SPOT and an equal-notional SHORT on the HYPERLIQUID BTC perpetual to harvest Hyperliquid's funding premium. This deliberately differs from the factory's existing Binance-internal spot-perp carry (IntraBinanceSolSpotPerpFundingCarryDeltaNeutral1H) and the perp-perp funding-spread rotations: the short leg lives on HYPERLIQUID, NOT on the BINANCE USD-M perp that the abandon-cluster confirmed is over-exposure-corrupted. Both chosen venues size correctly (BINANCE_SPOT is a CASH account; HL sizing is clean at ~64-90% avg_position_pct per prior runs), so this construction is immune to the USD-M position_size->order-qty defect that ruined the directional perp runs. HL funding accrues HOURLY (vs Binance's 8h), and crypto perps structurally trade at a premium (longs pay shorts) in neutral-to-bull regimes, so a perp short collects funding. The strategy only deploys when funding is ELEVATED enough that accrued carry over the expected hold beats the ~0.29% round-trip fee, keeping it firmly fee-positive. Few parameters (funding entry threshold, funding exit floor, basis stop, per-leg notional) keep overfitting risk low, addressing the recurring overfit failures. The 1H funding-check cadence accumulates many more trades than a daily-cadence HL strategy, sidestepping the HL data-sparsity wall that abandoned the daily Donchian/funding-reversion runs.

Implements the hypothesis exactly: real spot long on BINANCE_SPOT + short on the HYPERLIQUID perp, deliberately avoiding the over-exposure-corrupted Binance USD-M perp leg. Routing the long leg to BINANCE_SPOT (CASH account) makes get_account_equity() read true deployable USDT and keeps leverage at 1.0 (no unused-leverage rejection), while HL sizing is clean — both immune to the USD-M position_size->order-qty defect. The funding edge is harvested by holding the short HL perp; per the data spec the engine credits funding only as a signal driver, so the strategy times WHEN to hold the hedged pair. Critically, the entry gate is RELATIVE (rolling z-score on smoothed funding) rather than an absolute threshold: prior cross-venue carries produced ZERO trades twice because an absolute funding level is uncalibratable blind to the sandbox's funding scale (real ~1.25e-5/hr vs synthetic ~1e-4/hr); the z-score gate fires in any regime/data scale while still only deploying on genuinely elevated POSITIVE funding (positive carry). The hypothesis's distinguishing basis_stop is implemented as a directional delta-neutral guard — it unwinds both legs when the cross-venue basis moves adversely for the short-perp/long-spot pair, protecting mark-to-market without firing on benign entry-time basis offsets. min/max hold guards amortise the ~0.29% round-trip fee and cap stale carries. Parameter count is small (entry_z, exit_z, funding_lookback_hours, basis_stop_bps, target_notional_pct + structural knobs) to address recurring overfit failures.

No edge after two-venue costs — the third cross-venue BTC funding-carry construction this batch to fail the same way, and not fixable by optimization or code iteration. The strategy loses money in every month of the sample (Nov-2025 to Jun-2026, -0.15% to -4.14%): total_return -6.83%, Sharpe -2.36, Sortino -1.34, profit_factor 0.70, expectancy -$194/trade, PSR 0.0. The construction is correct and achieves its stated design goal — it dodges the USD-M position-size defect by using a BINANCE_SPOT CASH long + HL perp short, both legs size cleanly (avg_position 22%), and the book is genuinely delta-neutral (beta 0.005, balanced 33L/33S) — so this is an honest no-edge result, not a sizing bug. Fees dominate: commission_pct_of_gross 13.87% ($4.2k commission + $1.7k impact on 66 trades), so the HL hourly funding premium the short leg collects on BTC, net of the ~0.29% two-venue round-trip, is negative. There is a suspected leg-netting blip on the final bar (2026-06-23 -414% daily, return_kurtosis 122, skew -10.9) that is implausible for a delta-neutral pair (symptom noted, not independently verified; end_unrealized is only -0.45% so it does not drive the headline) — the decisive finding is the consistent fee-dominated monthly bleed. NOT iterate: the design is already right and the loss is structural (cost > edge), not a tunable parameter — entry_z/exit_z changes just trade less of the same negative-after-cost carry; this is also NOT a funding-uncreditable false negative (funding IS credited/accrued). Possible Research-Lead reframe (revise_hypothesis): retry the same delta-neutral cross-venue mechanism on a higher-funding-skew instrument (e.g. a high-funding altcoin perp) where the premium exceeds two-venue fees. FAILURE PATTERN: cross-venue BTC funding carry (long Binance spot / short HL perp) is genuinely delta-neutral and correctly sized but structurally unprofitable — BTC's HL funding premium is smaller than the ~0.29% two-venue round-trip cost, so the book bleeds every month (PF 0.70, Sharpe -2.36) with fees ~14% of gross; this is the third BTC cross-venue funding-carry construction (z-scored perp-perp, APR perp-perp spread, and now spot-HL) to die on the identical thin-premium-vs-fees wall, confirming BTC is a dead target for cross-venue funding carry.