EthSolPositiveFundingSpotPerpCarryDeltaNeutralBasket8H

ETH+SOL Positive-Funding Spot-Perp Carry Basket, Delta-Neutral (Cross-Venue: long BINANCE_SPOT + short BINANCE USD-M perp, per-leg funding-gated, held for days, 3-parameter)

A DELTA-NEUTRAL, cross-venue funding-carry basket over two independently-gated legs: ETH and SOL. For each asset, when the 8h perpetual funding rate is persistently and richly POSITIVE (longs paying shorts), the strategy simultaneously buys the asset on BINANCE_SPOT and shorts the equal-notional USD-M perpetual on BINANCE, collecting funding while carrying zero net directional (delta) exposure. This is deliberately the POSITIVE-funding mirror of the promoted long-only NEGATIVE-funding basket (which holds long-perp only) and a distinct expression from the single-asset BTC positive-funding carry: it (a) adds SOL, whose funding runs structurally hotter and more persistent than BTC/ETH, and (b) gates each leg on its OWN funding richness rather than a fixed BTC threshold, so the book only holds carry that clears the round-trip cost. Rationale for choosing this: it fills the two most under-represented portfolio buckets at once — cross_venue (7.1% vs 15% target) and market-neutral/hedged direction (diversifying away from the 85% long-only concentration) — while reusing the ONLY carry mechanism in the portfolio that has actually reached paper (Sharpe 8.58 BTC positive carry, 1.97 negative-funding basket). It is NOT a rank-and-rotate dispersion basket (L12): each leg is an independent, low-turnover, held-for-days delta-neutral carry, not a relative-strength rotation.

Iteration 2 addresses the Layer-4 QA rejection, which was purely an economic-viability calibration issue (every earlier layer passed). Root cause: entry_gate defaulted to 0.0001/8h (~3 bps/day), ~5x below the L13 ~15 bps/day carry-viability floor and well under the ~0.30% two-legged round-trip cost (spot 0.20% + perp 0.10%), so opened carry netted negative (total_return -2.02%, avg_trade_return_pct ~0.0007%). Minimal fix, no logic/import/structure changes: raise entry_gate default to 0.0006/8h (0.18%/day = 18 bps/day), which clears the 15 bps floor with margin and, because fees are one-time while funding accrues per settlement, a 3-day hold earns ~54 bps gross vs ~30 bps round-trip (net positive) and multi-day holds clear it decisively and beat the 0.15% avg-trade floor. exit_gate raised from 0.0 to 0.0001/8h (3 bps/day) so pairs are closed as carry decays rather than held near-zero, which is what let residual basis drift accumulate into the -10.3% end unrealized. The persistence gate (min_funding_pts=2 over a 3-point/~24h lookback) is retained so entries only fire after funding has been richly positive for ~24h, avoiding one-bar whipsaws. SOL 8h funding routinely exceeds 0.0006 across the 1307-day window, so the basket still produces ample trades (no no-trades/verification_loop regression) while now holding only carry that clears costs. Delta-neutrality (equal-notional legs, naked-leg guard) is unchanged; the near-symmetric win/loss confirmed gross neutrality already held, so the fix targets the entry economics, not the hedge.

fee/edge failure (L22): avg_trade_return_pct is -0.0138% — negative and far below the 0.15% futures floor — with profit_factor 1.055 and total_return -1.30%, while the headline Sharpe 3.59 masks it (the classic high-Sharpe/below-cost-per-trade trap that killed the near-identical ETH Cross-Venue Positive-Funding Carry after 3 wasted iterations). Root cause is structural, not a code bug: the two-legged delta-neutral construction pays ~0.30% round-trip (spot ~0.20% + perp ~0.10%) to synthesize neutrality, and on ETH/SOL the net positive funding captured per multi-day hold does not clear that double cost, so the average trade is a percent-loss even though a few large-notional holds keep USD expectancy positive. The developer already raised the entry gate once (0.0001 -> 0.0006/8h) and the per-trade edge is still negative, so further threshold tuning cannot help. This is NOT a funding-crediting problem (funding is credited; the mechanism is proven in promoted siblings) — it is that the positive-funding delta-neutral expression is inherently fee-disadvantaged versus single-leg funding capture, so a different asset would not fix the cost structure. Abandon at review rather than spend a 2-hour optimization on a below-cost edge.

Delta-neutral, cross-venue positive-funding carry basket over ETH and SOL. For each asset independently, when the 8h USD-M perpetual funding rate has been persistently and richly positive (trailing mean over ~24h >= entry_gate), it LONGs the asset on BINANCE_SPOT and SHORTs the equal-notional USD-M perp on BINANCE, collecting funding on the short with zero net delta. Each pair is closed when carry decays below exit_gate, on a 30-day timeout, or if a leg is ever left unhedged. Only the funding premium is the edge — no price/basis proxy substitution; a leg with no funding series stays flat.

Verification failed (Layer 4 — QA review): - entry_gate defaults to 0.0001 (0.01%/8h = ~3 bps/day), 5x below the ~15 bps/day carry-viability floor (L13: funding viable ONLY when median daily accrual > 15 bps; L1: cross-venue delta-neutral only when captured spread exceeds ~0.30% round-trip). The hypothesis claims the gate ensures 'the book only holds carry that clears the round-trip cost' — at 3 bps/day it does not. L6 arithmetic: ~3 bps/day × 15-day hold = ~45 bps gross vs ~0.30% two-legged round-trip (spot 0.20% + perp 0.10%); entries at the minimum gate net negative. Full-window sandbox (1307 days, 164 trades, metrics_reliable) confirms: total_return -2.02%, avg_trade_return_pct +0.000665% (~zero, far below 0.15% floor), end_unrealized_pct -10.30. Same defect as the SOL/DOGE/AVAX sibling, applied consistently. - Sharpe 4.48 (CI [3.90, 5.17], PSR 1.0) is high but coexists with NEGATIVE total_return (-2.02%) and -10.3% end unrealized — a metrics-shape artifact (many tiny positive funding accruals, positive skew 4.27, kurtosis 41.85) masking a net-losing book, not a promotable edge. - end_unrealized_pct -10.30 on a supposedly delta-neutral book suggests residual basis/carry exposure. largest_win/loss are near-symmetric (+31.6k/-28.5k) so gross delta-neutrality roughly holds, but confirm this is accumulated funding/basis on open pairs and not a sizing/hedge-ratio mismatch (equal-notional legs at prec 3/2 rounding).

Confirm end_unrealized captures BOTH legs (spot cash gain + perp margin loss); if perp-only, total_return understates the strategy.

Recurring large NEGATIVE end_unrealized_pct (-9.34% here, -10.30% at iter-1) on a supposedly delta-neutral book. A truly hedged pair should mark near zero, not -9%. Most likely a metric-presentation artifact — perp (MARGIN-leg) unrealized shown WITHOUT the offsetting spot (CASH-leg) gain over the ~13.8-day hold — consistent with net-positive realized PnL (PF 1.56) and near-symmetric largest_win/loss (+$20.3k/-$15.0k). But must be confirmed it is not a real unhedged residual, since it is the sole reason total_return is negative (-0.49%) despite +$680/trade realized expectancy.

Evaluate on total_return/Sharpe/funding over the full window; abandon at BACKTEST_REVIEW if total_return stays negative once end_unrealized is correctly attributed.

avg_trade_return_pct (+0.0333%) is now positive (fix worked) but below the 0.15% floor — which does not apply to a delta-neutral pair book (per-leg percent dominated by hedged-away price move). Meaningful measures: total_return, Sharpe (11.95, CI [9.68,14.54]), net funding captured. Sample small (26 trades/192 days).

On the full run, assert funding cash flow on the short perp is positive while funding>0; if mis-signed, that's an engine fix, not a strategy change.

Verify short-perp funding accrual SIGN. This positive-funding SHORT-perp construction is untested vs the promoted NEGATIVE-funding LONG-perp carry. The marginal net result means the analyst should confirm the engine CREDITS positive funding to the held short (a short RECEIVES funding when funding>0).

Proven mechanism family (funding carry) with promoted siblings; funding is credited so the carry is real in-backtest

Exits and hedging function correctly this time: avg_holding 11d within the 30d cap, max_drawdown 1.30%, annualized_vol 0.68%, commissions only 1.52% of gross

benchmark_meaningful=false correctly flags market-neutral

avg_trade_return_pct is NEGATIVE (-0.0138%) — below the 0.15% futures floor and below zero; the average trade loses money on a percent-of-notional basis

profit_factor 1.055 sits in the fee-edge dead zone; total_return -1.30%

Sharpe 3.59 is misleading — high Sharpe on a below-cost per-trade edge is fee-fragility, not signal quality (L22)

Two-legged spot+perp construction pays ~0.30% round-trip to synthesize delta-neutrality, which structurally exceeds the ETH/SOL positive-funding carry captured per hold

Entry gate already raised once (0.0001 -> 0.0006/8h) and the per-trade edge is still negative — threshold tuning is exhausted

EthSolPositiveFundingSpotPerpCarryDeltaNeutralBasket8H tried to mirror the factory's proven funding-carry family on the positive side, adding SOL for its hotter funding and gating each leg on its own richness, to fill the under-represented cross-venue and market-neutral buckets. The engineering was clean this time — perfectly delta-neutral, funding genuinely credited, tidy 11-day holds, tiny drawdown and low fees — but the economics failed structurally: the two-legged spot+perp hedge pays ~0.30% round-trip to synthesize neutrality, and ETH/SOL positive funding per hold doesn't clear it, leaving avg_trade_return_pct negative while a few large holds kept USD expectancy barely positive and inflated Sharpe to 3.59. The reviewer abandoned it as a fee/edge failure, noting the entry gate had already been raised and threshold tuning was exhausted, and that this was the same high-Sharpe/below-cost trap that had wasted iterations on the near-identical ETH positive carry. The mechanism is proven for single-leg funding capture, but the delta-neutral positive-funding expression is inherently cost-disadvantaged — no asset swap fixes it.

A high Sharpe on a below-cost per-trade edge is fee-fragility, not signal quality: the positive-funding delta-neutral expression is structurally fee-disadvantaged versus single-leg funding capture because synthesizing neutrality costs a ~0.30% round-trip that ETH/SOL positive funding doesn't clear — a different asset wouldn't fix the cost structure, and this repeats the trap that killed the near-identical ETH cross-venue positive carry.

The backtest reviewer returned 'abandon' at the pre-optimization gate for a fee/edge failure: the two-legged spot+perp construction pays ~0.30% round-trip to synthesize delta-neutrality, and on ETH/SOL the net positive funding captured per multi-day hold does not clear that double cost, so the average trade loses on a percent-of-notional basis — the entry gate had already been raised once (0.0001→0.0006/8h) and the per-trade edge stayed negative, making further threshold tuning futile.

A delta-neutral, cross-venue positive-funding carry basket over two independently-gated legs (ETH and SOL): for each asset, when 8h perp funding ran persistently and richly positive, it bought spot on BINANCE_SPOT and shorted the equal-notional USD-M perp on BINANCE to collect funding with zero net delta, holding for days and gating each leg on its own funding richness — 3 parameters, the positive-funding mirror of the promoted long-only negative-funding basket.

With full data (60 trades, funding credited, metrics_reliable=true) the hedging and exits worked cleanly — genuinely delta-neutral (beta -0.0001), avg holding ~11 days within the 30-day cap, max drawdown 1.30%, annualized vol 0.68%, commissions only 1.52% of gross — but the per-trade edge was negative: avg_trade_return_pct -0.0138% (below zero and far below the 0.15% futures floor), profit factor 1.055, total return -1.30%, with a headline Sharpe of 3.59 that masked the below-cost edge. No optimization stage ran.