BnbFundingCarryDeltaNeutralPercentile8H

BNB Positive-Funding Spot-Perp Cash-and-Carry, Single-Asset Delta-Neutral, Percentile-Triggered (Cross-Venue: long BNBUSDT.BINANCE_SPOT + short BNBUSDT.BINANCE USD-M perp, held for days while funding sits in its own upper distribution, 3-parameter)

A DELTA-NEUTRAL, cross-venue funding cash-and-carry on BNB — the one top-5-liquidity name NOT yet covered by the factory's carry book (BTC single, AVAX single, ETH+SOL basket are already in-pipeline). BNB has active, deep BINANCE_SPOT and USD-M perp markets with full funding history, and its funding is event-driven (Binance launchpad/BNB-chain catalysts) so it spikes into richly positive territory in bursts that persist for days. When BNB 8h funding is richly positive (longs paying shorts), the strategy BUYS BNB on BINANCE_SPOT and SHORTS the equal-notional BNBUSDT USD-M perp, carrying zero net delta while collecting funding every 8h, held for DAYS so the funding stream dominates the two-leg round-trip cost. The key differentiator from the fixed-threshold sibling carries is a SELF-ADAPTIVE PERCENTILE trigger: it enters only when funding exceeds both an absolute floor AND its own trailing 90-day upper distribution, which (a) adapts to BNB's regime instead of a hand-tuned constant (reduces overfit — the #1 killer at 203x), and (b) guarantees the position is only on during genuinely rich-carry episodes. It is single-asset (no basket rebalance churn — only 2 fills per round trip, sidestepping L12) and a directional funding cash flow, not a spread-arb (L1) or a funding-fade (L13). It fills the two most under-represented buckets at once: cross_venue (7.3% vs 15% target) and market-neutral direction (diversifying from the 85.3% long-only concentration), and reuses the ONLY portfolio mechanism that has reached paper (BTC positive-funding carry, Sharpe 8.58).

Implements the hypothesis by reusing the proven single-asset carry machinery (funding parse, searchsorted lookups, timestamp-aligned two-leg management, naked-leg guard, fixed capital_frac sizing) and adding the stated differentiator: a self-adaptive percentile entry. The gate requires funding to clear BOTH an absolute floor AND its own trailing-90-day upper-distribution quantile. The floor (0.0003/8h) guarantees viability against the ~0.30% cross-venue round trip over multi-day holds (L1/L13), directly fixing the failure mode that sank the mild-gate basket (3 bps/day was below fee cost); the percentile adapts to BNB's event-driven funding regime rather than a hand-tuned constant, cutting overfit (the #1 killer at 203x). It is single-asset (2 fills per round trip, no basket rebalance churn, sidestepping L12) and captures a directional funding cash flow, not a spread-arb (L1) or funding-fade (L13). Venue routing is correct: primary is the BNB USD-M perp (MARGIN account gives a stable equity read and is the funding-accounting/short leg); the spot long is the cross-venue hedge on BINANCE_SPOT (CASH, long-only — the strategy only ever BUYS spot, never shorts it). BNB is a top-5-liquidity name with deep spot+perp markets and full funding history, so both legs are collectible (no L4 missing-leg risk). leverage is 1.0 and referenced in the notional (no leverage_set_but_unused). I dry-ran the combined floor+percentile gate on synthetic BNB-like funding (baseline ~0.8 bp/8h with multi-day spikes): it produced 31 entry-eligible bars over the rich episodes, so the sandbox will generate trades (no verification_loop / no-trades risk). The continuous signal is the trailing funding mean, varying every bar; entry/exit decisions are thresholded against that same per-8h unit. Fills the under-represented cross_venue and market-neutral buckets with low turnover.

fee/edge failure (L22): avg_trade_return_pct is 0.0155% — ~10x below the 0.15% futures floor — with profit_factor 1.069 and total_return -3.65%, while Sharpe 1.77 masks it (the same high-Sharpe/below-cost trap as the abandoned AVAX and ETH+SOL positive-funding carries). impact_cost_pct 28.1% and capacity_usd $1.27M add further drag/scale limits. Root cause is structural and identical across this family: the two-legged spot+perp construction pays ~0.30% cross-venue round-trip to synthesize delta-neutrality, and BNB positive funding — even gated to both an absolute floor and its own 90-day upper-distribution percentile — does not net clear that cost, leaving a per-trade edge of 0.0155%. The percentile trigger genuinely reduces overfit but cannot create edge where cost exceeds the funding captured, so no parameter tuning helps. This is NOT a funding-crediting problem (funding is credited; the mechanism is proven in the promoted BTC sibling) — the positive-funding TWO-LEGGED expression is inherently fee-disadvantaged. META-NOTE FOR RESEARCH LEAD: this exact construction (long spot + short perp positive-funding delta-neutral carry) has now failed the fee floor on AVAX (avg_trade_return_pct -0.0058%), ETH+SOL (-0.0138%), and BNB (0.0155%) for the same structural reason — the second (spot) leg's fees swamp the carry. Retire this positive-funding two-legged carry family rather than re-pointing it at more alts; the promoted carry sibling works because it captures funding with a SINGLE leg (long-perp on negative funding), avoiding the second leg's round-trip. If carry is still wanted in the cross-venue/market-neutral buckets, pursue single-leg funding capture, not two-legged synthetic delta-neutral.

Single-asset, delta-neutral, cross-venue positive-funding cash-and-carry on BNB with a self-adaptive percentile trigger. On 8-hour bars it reads BNB perp funding from supplementary data (per-instrument series with fallback to the primary accounting series). It opens a carry pair — BUY BNBUSDT.BINANCE_SPOT + SHORT equal-notional BNBUSDT.BINANCE USD-M perp — only when the trailing 3-point funding mean exceeds BOTH an absolute viability floor (entry_floor = 0.03%/8h ≈ 9 bps/day, enough that a multi-day hold clears the ~0.30% two-leg round trip) AND the 85th percentile of funding's own trailing-90-day distribution. The short perp collects funding every 8h while the long spot cancels price risk (delta-neutral). The pair is held for days until funding normalizes (trailing mean ≤ exit_gate = 0.01%/8h) or a 40-day timeout. A naked-leg guard flattens both legs if one ever exists without the other. Funding is parsed once into sorted numpy arrays; per-bar work is O(log n) searchsorted plus small-window mean/percentile. Three tunable knobs: entry_floor, entry_pctile, capital_frac.

Backtest engineer/analyst, before promoting ANY carry: (1) confirm end_unrealized sums BOTH legs; (2) assert short-perp funding is POSITIVE while funding>0. Until settled, the family's backtests can't be trusted.

CROSS-CUTTING, NOW 5 INSTANCES — MUST BE RESOLVED BEFORE ANY OF THIS FAMILY PROMOTES. Every spot-perp positive-funding carry reviewed (SOL/DOGE/AVAX basket, ETH+SOL, single AVAX, now single BNB) prints a NEGATIVE total_return driven by a large NEGATIVE end_unrealized_pct (-9.32% here; -6.46%, -9.34%, -10.30% on siblings) on a supposedly delta-neutral book, while REALIZED expectancy is positive (+$194/trade, PF 1.119, Sharpe 3.38, avg_trade_return_pct +0.0148%). Either (a) end_unrealized is perp-leg-only (metric artifact understating total_return) or (b) the engine isn't crediting positive funding to the held short. Near-symmetric largest_win/loss (+$56.7k/-$51.9k) and positive realized expectancy point to (a), but five independent instances mean it can't be deferred.

Analyst: confirm the floor+percentile keeps net carry positive on the full window; consider nudging entry_floor to 0.0005 for margin. avg_trade_return_pct is a per-leg artifact — gate on total_return/Sharpe/funding, not the 0.15% directional floor.

entry_floor 0.0003/8h (9 bps/day) is below the ~15 bps/day L13 line I enforced on siblings, but acceptable here: 9 bps/day x ~6.1-day hold ~= 54 bps gross vs ~30 bps round-trip = net positive; the 85th-pctile gate raises the effective threshold in rich episodes; and the sandbox shows a POSITIVE per-trade return (unlike the 3 bps/day cases). Margin is thinner than the 0.0005 floor elsewhere.

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

Percentile trigger is a genuine anti-overfit improvement; exits/hedging work (avg_holding 4.7d within cap, commissions 5.24% of gross)

benchmark_meaningful=false correctly flags market-neutral

avg_trade_return_pct 0.0155% — ~10x below the 0.15% futures floor (L22)

profit_factor 1.069 in the fee-edge dead zone; total_return -3.65%

Sharpe 1.77 is misleading — high Sharpe on a below-cost per-trade edge is fee-fragility, not signal quality

impact_cost_pct 28.1% and capacity_usd $1.27M — meaningful impact drag and limited scale

Same structural failure as the AVAX and ETH+SOL siblings: two-legged ~0.30% round-trip exceeds the BNB funding captured; percentile trigger cannot fix a cost>edge problem

BnbFundingCarryDeltaNeutralPercentile8H was the fourth attempt at the cross-venue positive-funding cash-and-carry, pointing it at BNB (the last uncovered top-5 name) and adding a genuinely anti-overfit self-adaptive percentile trigger on top of the absolute funding floor. The mechanics were sound — perfectly delta-neutral, funding credited, tidy multi-day holds — and the percentile gate was a real improvement, but the economics failed identically to its siblings: a per-trade return of 0.0155% sat ~10x below the fee floor, profit factor languished at 1.069, and impact ate 28% of gross, with Sharpe 1.77 disguising the below-cost edge. The reviewer abandoned it on the first iteration and, notably, escalated a meta-note: having now failed the fee floor on AVAX, ETH+SOL, and BNB for the same reason — the spot leg's fees swamp the carry — the entire two-legged positive-funding carry family should be retired in favor of single-leg funding capture, which is why the promoted BTC/negative-funding sibling works.

This exact long-spot/short-perp positive-funding delta-neutral carry has now failed the fee floor on AVAX, ETH+SOL, and BNB for the same structural reason — the second (spot) leg's round-trip fees swamp the captured funding — so the whole two-legged positive-funding family should be retired rather than re-pointed at more alts; the promoted carry sibling works precisely because it captures funding with a single leg (long-perp on negative funding), avoiding the second leg's cost.

The backtest reviewer returned 'abandon' at the pre-optimization gate for a fee/edge failure: the two-legged spot+perp construction pays ~0.30% cross-venue round-trip to synthesize neutrality and BNB positive funding — even gated to both an absolute floor and its own percentile — does not net clear it, so the average trade earns below the fee floor; the percentile trigger reduces overfit but cannot create edge where cost exceeds carry, and no parameter tuning helps.

A delta-neutral, cross-venue positive-funding cash-and-carry on BNB — long BNBUSDT.BINANCE_SPOT against a short equal-notional USD-M perp — collecting funding while carrying zero net delta, held for days, with a self-adaptive trigger that entered only when funding exceeded both an absolute viability floor AND its own trailing-90-day upper-distribution percentile to reduce overfit.

With full data (78 trades, funding credited, metrics_reliable=true) the hedge worked (beta -0.0001, avg holding ~4.7 days, commissions 5.24% of gross) but the per-trade edge was below cost: avg_trade_return_pct 0.0155% (~10x below the 0.15% futures floor), profit factor 1.069, total return -3.65%, with impact_cost_pct 28.1% and capacity of $1.27M — while the headline Sharpe of 1.77 masked the below-cost edge. No optimization stage ran.