BtcCashAndCarryFundingArbExtremeTriggeredNeutral

BTC Cash-and-Carry Funding Rate Arbitrage (Spot Long + Perp Short, Extreme-Funding Triggered)

Market-neutral cash-and-carry funding rate arbitrage capturing positive funding payments from perpetual shorts while hedging delta with a matched spot long. When BTCUSDT perpetual 8h funding rate exceeds an extreme threshold (>=0.05% per 8h, equivalent to >=54.75% APR), simultaneously open LONG BTCUSDT.BINANCE_SPOT and SHORT BTCUSDT.BINANCE perpetual in EQUAL USD notional (delta-neutral). Each 8h funding settlement the short perp leg RECEIVES the funding payment from longs; the spot leg eliminates directional BTC exposure. Position is held until funding normalizes below a cooldown threshold (<0.01% per 8h) or maximum hold time reached. This is a structural, mechanically-driven cash flow strategy that does NOT depend on price prediction. The 'extreme funding' regime occurs ~5-15% of the time in crypto bull runs and during volatility spikes — it is a real, documented persistent edge not arbitraged away because of capital constraints (requires simultaneous deployment on two venues with cross-margin coordination). Uses 4h bars (NOT 8h, which has had data availability issues in recent failures) so the funding settlement timestamp (every 8h) is captured cleanly across 2 bars.

Implements the documented cash-and-carry funding edge: in bull/vol-spike regimes (~5-15% of the time) perp funding goes extreme and persists because the arb requires simultaneous two-venue capital deployment with cross-margin coordination, so it is not fully arbitraged away. It is a mechanical cash-flow harvest (the mechanism family this factory advances through the pipeline), not a price-prediction strategy. Leg assignment respects venue constraints: spot is CASH/long-only so it is the BUY-only primary, the shortable USD-M perp is the extra leg; equal USD notional makes the book delta-neutral so funding accrual is the sole PnL source. Extreme entry threshold (0.05%/8h) ensures cumulative funding far exceeds the ~0.2% spot + ~0.1% perp round-trip fees over a multi-cycle hold. Deliberately 3 edge parameters to resist walk-forward overfitting. leverage=1.0 (spot cap) and sizing uses capital_frac not leverage, so no unused-leverage gate triggers.

BTC is structurally the wrong target for extreme-funding cash-and-carry, and the mechanism is already validated and PROMOTED on ETH — so the family's value is captured and this BTC instance should be retired rather than iterated. Despite a high full-period Sharpe (13.6, DSR 1.0, is_significant TRUE), the decisive generalization gates fail: the holdout produced ZERO trades (holdout_sharpe 0.0, ratio 0.0) and the walk-forward is_overfitted=TRUE with OOS windows [0.0, 0.0, 13.20] — two of three zero-trade, the OOS Sharpe (4.40) propped up by a single 2020/2021 window. DSR=1.0 is a full-period statistic that does not reflect this clustering; the holdout and is_overfitted gates catch it and both fail. Root cause: BTC extreme funding (>=0.0005/8h) occurs only in bull-mania regimes (2020/2021, brief 2024) and is absent in 2022/2023/2025/2026, so recent walk-forward/holdout windows are empty — the optimizer again over-selected toward sparsity (funding_entry 0.0005->0.000749, 12 trades). This is the IDENTICAL failure to the BTC 8H sibling (13a0b247) and is corroborated by the multi-asset basket review (b374fac1) where even spreading across 5 majors left 2025/2026 empty because major-coin extreme funding is correlated and regime-clustered. Not iterate: (1) the lower-threshold density fix is ALREADY being tested on the redundant BTC 8H sibling (13a0b247, attempt 1 of 2) — iterating this 4H duplicate runs the same experiment twice (I flagged the 8H/4H redundancy at pre-optimization); (2) BTC's recent funding is structurally compressed, so a fee-clearing lower threshold is unlikely to populate the empty recent windows; (3) the explicit redirect target named in the BTC-sibling feedback — the ETH spot-perp sibling (2c9b5c9e) — has now PROMOTED (DSR 0.99, holdout ratio 1.198, OOS>IS), so the higher-funding-instrument escalation already succeeded. FAILURE PATTERN: same-venue extreme-funding cash-and-carry clears fees and is genuinely market-neutral, but on BTC the fee-clearing extreme-funding gate fires only in 2020/2021-style manias, producing zero-trade recent holdout/walk-forward windows (is_overfitted=TRUE, holdout 0.0) regardless of bar cadence (4H or 8H) — a high full-period DSR cannot rescue it. The mechanism belongs on richer-funding instruments (ETH, promoted); BTC is a dead target for it.

Market-neutral cash-and-carry funding-rate arbitrage on BTC. Long leg = BTCUSDT.BINANCE_SPOT (CASH, long-only, primary); short leg = BTCUSDT.BINANCE USD-M perpetual (extra). The signal is the most-recent settled Binance BTCUSDT 8h funding rate, looked up from the funding_rates supplementary series (flat {ns: rate}) at each 4-HOUR bar timestamp (smooth_window=1 keeps it the raw latest rate so genuine spikes are not diluted). When funding is EXTREME (>= funding_entry, default 0.05%/8h ~= 54.75% APR) the strategy opens a delta-neutral book: BUY spot + SELL an equal-USD-notional perp. Each 8h settlement the perp short receives the funding payment while the matched spot long cancels BTC price exposure. It flattens BOTH legs when funding normalizes (<= funding_exit, default 0.01%/8h) or a frozen max-hold / combined-PnL kill switch fires. 4-HOUR bars cleanly straddle the 8h settlement across 2 bars and avoid the 8h-bar data gaps. 3 edge parameters (funding_entry, funding_exit, capital_frac); smoothing and backstops frozen. If funding data is absent the signal stays 0 and no position opens -- no price-pattern fallback.

Funding is substantially credited (+$22,434) and dominates the small basis residual (-$5,677), for +$16,757 net realized with 0 open legs — the cash-and-carry mechanism works and the hedge is neutral over holds (2021 pair: spot +13,887 / perp -8,728, price residual -$272, funding +$5,431 = the gain).

Cleanest funding-carry profile of the session: return_skew -0.70, kurtosis 5.1, max_drawdown 4.76% — the high 'extreme funding' entry threshold keeps holds short/selective and basis risk low.

Realized risk-adjusted metrics strong and plausible (not artifact-tier): Sharpe 7.54 from genuinely low delta-neutral vol (1.05% annualized), sharpe_ci_low 5.37, PSR 1.0, Sortino 11.5; annual returns all positive (2020/2021/2024); capacity $22M; low fee drag (4.5% of gross).

Headline total_return -2.23% and end_unrealized_pct -10.6% contradict the +$16,757 realized trade sum with 0 open legs — an MTM/accounting artifact to verify; judge on realized (positive).

Sparse and clustered: 40 entries all in 2020/2021/2024 (extreme-funding episodes), ZERO in 2022/2023/2025/2026 — walk-forward OOS windows risk near-empty trade counts (degenerate walk-forward / DSR risk on only 40 trades).

Redundant with the BTC spot-perp funding carry already advancing to optimization (hypothesis 13a0b247, 8H) — same mechanism and asset on a different cadence; the two will be highly correlated.

+$16,757; all annual returns positive

positive

Sharpe 7.54, sharpe_ci_low 5.37, PSR 1.0, max_dd 4.76%

positive & significant

+$22,434 vs -$5,677 price residual

funding dominates

40 trades, all 2020/2021/2024, none 2022/2023/2025/2026

distributed for clean walk-forward

Mechanism is real and credited (verified pre-opt: funding +$22,434 vs -$5,677 basis residual, net positive, cleanest distribution of the session — skew -0.70, kurtosis 5.1, max_dd 4.76%).

Full-period optimized Sharpe is high and statistically clears the multiple-testing bar in isolation: DSR 1.0, is_significant TRUE, sharpe_ci_low 10.69, optimized Sharpe 13.6 > expected-max 5.14.

Sensitivity essentially clean (1 benign funding_entry cliff that stays positive); the mechanism itself is sound — it is simply on the wrong asset.

Holdout FAILED: holdout_sharpe 0.0, ratio 0.0 — the recent holdout window had ZERO extreme-funding trades.

Walk-forward is_overfitted=TRUE: avg IS 14.36 vs avg OOS 4.40, OOS windows [0.0, 0.0, 13.20] — two of three are zero-trade, generalization propped by a single 2020/2021 window.

Optimizer again over-selected toward sparsity (funding_entry 0.0005->0.000749, 12 trades), all clustered in 2020/2021/2024 with nothing in 2022/2023/2025/2026 — BTC extreme funding is structurally confined to bull manias.

pbo 0.4801 is borderline (just under 0.5). DSR 1.0 is a full-period artifact that does not reflect the zero-trade-window clustering — the holdout and is_overfitted gates do, and they fail.

Identical failure to the BTC 8H sibling (13a0b247) and confirmed by the multi-asset basket review — BTC is the wrong target for this mechanism.

2 of 3 are zero-trade [0.0, 0.0, 13.20]

all windows populated

BtcCashAndCarryFundingArbExtremeTriggeredNeutral ran the textbook long-spot/short-perp carry on BTC, gated to only the most extreme funding so the perp short is well-paid, and it produced the session's cleanest funding-carry profile — funding credited +$22,434, +$16,757 net realized, Sharpe 7.54, 4.76% drawdown — earning a full optimization. But the edge was regime-clustered: all 40 trades fell in 2020/2021/2024, so the optimized run failed its holdout with zero trades and was walk-forward-overfitted despite a 13.6 full-period Sharpe. The analyst ruled BTC a structurally dead target for the mechanism — identical to the abandoned 8H sibling — while noting the ETH sibling had already promoted, so the family's value was captured elsewhere. It ended after one iteration as abandoned, reaching optimization and analysis but never risk review.

Extreme-funding cash-and-carry is a genuinely market-neutral, fee-clearing edge but BTC is the wrong target regardless of bar cadence (4H or 8H): its fee-clearing funding spikes only occur in bull manias, leaving recent walk-forward/holdout windows empty (is_overfitted=true, holdout 0.0) — the family's value belongs on richer-funding instruments (ETH, now promoted), and a flattering full-period DSR is a statistic that masks regime-clustered sparsity.

It passed the pre-optimization backtest-review gate (verdict: optimize) but was abandoned at the post-optimization analyst gate (verdict: abandon): BTC extreme funding fires only in 2020/2021-style bull manias and is absent in 2022/2023/2025/2026, so the recent holdout and two of three OOS windows had zero trades — a high full-period DSR cannot rescue zero-trade-window clustering — and the mechanism is already validated and PROMOTED on ETH, so this BTC instance was retired rather than iterated, never reaching risk review.

A market-neutral cash-and-carry funding-rate arbitrage on BTC — when the BTCUSDT perp 8h funding rate exceeded an extreme threshold (≥0.05%/8h, ~54.75% APR), simultaneously longing BTCUSDT.BINANCE_SPOT and shorting the BTCUSDT.BINANCE perp in equal USD notional so the short perp collects the rich funding while spot hedges price, flattening on funding normalization, on 4H bars with just three edge parameters.

The baseline was the cleanest funding-carry of the session — 40 trades, funding credited +$22,434 dominating a -$5,677 basis residual for +$16,757 net realized, Sharpe 7.54 (CI [5.37, 10.28]), skew -0.70, kurtosis 5.1, max drawdown 4.76%, all annual returns positive (a -2.23% headline flagged as an MTM artifact); optimization then lifted the full-period Sharpe to 13.6 (DSR 1.0, significant) but failed generalization — holdout produced ZERO trades (holdout Sharpe 0.0, ratio 0.0) and walk-forward was overfitted (avg IS 14.36 vs OOS 4.40, windows [0.0, 0.0, 13.20]).