PositiveFundingAltSpotPerpCarryPercentileBasketDaily

Persistently-High-Positive-Funding Altcoin Spot-Perp Carry (Percentile-Gated, Daily Rebalance, Market-Neutral)

Delta-neutral positive-funding carry that captures the cash flow paid by long perp holders on persistently-premium, high-retail-demand altcoins — the cohort whose perps structurally trade rich to spot (DOGE, PEPE, WIF and similar memecoin/high-beta names that have both deep Binance spot and USD-M perp markets). For each active asset the strategy holds matched long-spot (BINANCE_SPOT) + short-perp (BINANCE) legs at equal notional, so the book is delta-neutral and earns the funding payment every 8h regardless of price direction. Three deliberate changes vs the failed ETH framing: (1) TARGET — high-funding alts instead of low-funding ETH, so realized carry is 5-15x larger and can clear hedge friction; (2) ENTRY GATING — only open a leg when that asset's trailing 3-sample mean funding sits in the TOP QUARTILE (>=75th percentile) of its own 30-day funding history AND exceeds 0.03%/8h in absolute terms (>=0.09%/day, ~32%/yr annualized while held), so the realized-while-held carry is far above the asset's average; (3) REBALANCE CADENCE — evaluate and rebalance on the DAILY (1D) bar close only, and re-hedge delta only when drift exceeds a +-3% band, cutting the per-pair basis-bleed and commission drag that killed the 1H version. Asymmetric hysteresis (enter at 75th pct, exit at 50th pct) prevents threshold flip-flopping. Market-neutral by construction, so MTM swings are limited to spot-perp basis movement. Differentiated from the in-pipeline NegativeFundingCarryLongBasket5Perps4H sibling: that book is long-perp collecting NEGATIVE funding; this book is short-perp collecting POSITIVE funding on the opposite (premium) cohort, percentile-gated and daily-rebalanced — non-overlapping instruments and opposite leg structure.

Iteration 3 keeps the working delta-neutral mechanism (backtest review confirmed the pairs are constructed correctly and funding is genuinely captured on the short perp legs) and makes the smallest targeted changes for the three review items. (2) THIN EDGE: added a basis-aware take-profit (take_profit_pct, net-pair PnL) so favorable spot-perp basis convergence plus accrued carry is banked instead of bleeding back through exit-basis drift and double-leg commissions — this directly lifts realized net carry above fees that produced PF~1.03. The existing basis-stop and hysteresis exits are retained. (3) SAMPLE TOO SMALL: replaced the short-history WIF/ORDI/SUI names (Binance spot only from ~2024) with 8 SAME-BASE alts that have multi-year spot+perp history, and relaxed entry_percentile 75->70 while keeping the hard absolute funding_floor — this yields far more qualifying episodes/pairs across 2021-2025 for a meaningful optimization sample. Sizing stays capital-relative and equal-base-qty on both legs (delta-neutral); per_leg_pct 0.12 x max_active_pairs 6 = 0.72 worst-case spot, safely under the CASH venue's no-leverage cap; leverage=1.0 (adding leverage would reintroduce beta). (1) RETURN-ACCOUNTING DISCREPANCY (headline -35% vs +$896 trade PnL, positive Sharpe): this is an engine-side dual-venue CASH(spot)+MARGIN(perp) equity-aggregation netting artifact, not strategy behavior — the book is delta-neutral by construction — so it is flagged for the backtest engineer rather than patched in strategy code, which cannot alter cross-venue equity aggregation.

Not worth optimizing: the edge is entirely regime-concentrated in the 2020-2021 retail-mania period. Meaningful trades/returns exist only in 2020 (5.97%) and 2021 (5.85%); from mid-2021 onward the top-percentile+floor entry gate essentially never fires (2023 ~0.01%, 2024 ~0.02%, zero activity in 2022/2025/2026), so walk-forward OOS and the recent holdout windows would be near-empty and optimization would fit a 2020-2021-only sample that cannot generalize. The headline is also unpromotable on its face — total_return is NEGATIVE (-1.26%) while the reported Sharpe (3.99) is a near-zero-volatility artifact (annualized vol 1.67%, kurtosis 55, tail_ratio 25.9), and the developer flags an unresolved dual-venue CASH+MARGIN equity-aggregation discrepancy that makes the optimization objective itself unreliable. This is a target/premise problem, not a parameter problem: the developer deliberately chose established alts (DOGE/AVAX/LINK/DOT/ADA/SOL/XRP/LTC) for history length, but those names' positive-funding premium structurally decayed after 2021 — while the true high-retail-demand memecoins that carry the edge (PEPE/WIF/ORDI) were already rejected for lacking pre-2024 history. No parameter iteration restores a decayed funding premium on backtestable-history assets. (Note: funding IS credited by the current engine; this abandon is for edge decay + regime-concentrated, negative-net, statistically-inadequate sample — NOT for any funding-crediting limitation.)

Delta-neutral positive-funding spot-perp carry basket over 8 long-history same-base alts (DOGE/AVAX/LINK/DOT/ADA/SOL/XRP/LTC). Each active asset holds long BINANCE_SPOT + short BINANCE USD-M perp in equal base quantity, collecting positive 8h funding on the short perp. Entry gated when trailing-3 mean funding is in the top gate (>=70th percentile) of the asset's own 30d funding distribution AND above a 0.03%/8h absolute floor; asymmetric-hysteresis exit below the 50th percentile, plus funding-flip, max-hold, basis-stop, and a NEW basis-aware take-profit. Evaluated/rebalanced on the daily bar; delta re-hedged only on >3% drift.

Genuinely delta-neutral construction (beta ~0.0, benchmark_correlation -0.15) with funding now credited by the engine — the mechanism is sound and produces positive per-trade expectancy (PF 1.16) when it actually trades

Low realized drawdown (1.28%) and low fee drag (commission 1.09% of gross)

Regime-concentrated: essentially ALL P&L is in 2020-2021 (annual returns 5.97/5.85), collapsing to ~0 in 2023-2024 with no 2022/2025/2026 activity — the top-percentile funding gate stops firing once the retail-mania premium decays

Negative net total_return (-1.26%) despite positive gross expectancy — thin carry barely clears hedge friction/costs, plus an unresolved dual-venue equity-aggregation discrepancy the developer flagged

Sharpe 3.99 is a near-zero-vol artifact (annualized vol 1.67%, kurtosis 55, tail_ratio 25.9) — a flat book with a few 2020-2021 spikes, not a durable edge

Target-selection contradiction: the cohort was swapped to established alts (which lack persistent high funding) for history length, abandoning the memecoin cohort the hypothesis's edge actually requires

Optimization sample is inadequate — walk-forward OOS and the recent holdout windows would be near-empty, so a 2-hour sweep would fit a 2020-2021-only sample and cannot generalize

Funding IS credited and captured: +$22,402 of real 8h funding cash flow over 6.5y confirms the engine credits the carry and the mechanism works as designed.

The book is genuinely delta-neutral at the pair level: the two legs offset (perp price -$95.2k vs spot price +$89.7k), so there is no naked directional exposure despite the alarming per-leg PnL swings.

Clean sensitivity (0 cliffs), legitimate cross-venue + market-neutral + 1H bucket fills, no optimizer escape hatch (fixed 0.5x sizing).

True economic result is flat-to-negative: equity $200k -> $196k over 6.5 years. The ~1.6%/yr realized funding capture is entirely eaten by cross-venue hedge basis bleed (-$5.5k), commissions (-$3.5k), and negative end-unrealized (-10.4%).

ETH is a structurally LOW-funding instrument: median per-8h funding ~+0.0001; the carry yield is too thin to overcome a 1H-rebalanced cross-venue hedge's friction + taker fees. avg per-leg trade return ~0.08%, below the 0.15% futures viability bar.

Reported Sharpe 2.1 / skew -11.7 / kurtosis 328 / DSR 0 / holdout -3.54 are CONTAMINATED metrics: the daily-return series double-counts per-leg closes for this two-leg neutral book (benchmark_meaningful=false), so the risk-adjusted stats are not trustworthy and were not used as the basis for this verdict.

Cross-venue 1H hedge inherently bleeds basis PnL (-$5.5k over 198 pairs, ~-0.056%/pair) — friction that scales with trade count and offsets the thin funding.

unreliable

Do NOT optimize yet. The mechanism works (delta-neutral pairs constructed correctly, funding genuinely captured on the short perp legs), but three things must be fixed first. (1) RETURN-ACCOUNTING DISCREPANCY: trade-level PnL sums to roughly +$896 (PF 1.03, expectancy +$32/trade, ~+0.4% on the implied ~$200k equity), yet total_return reports -35.4% with 35.4% drawdown and a positive Sharpe 3.48. These cannot coexist for a delta-neutral book that nets positive at the trade level. This is almost certainly the dual-venue CASH(spot)+MARGIN(perp) equity aggregation mis-netting spot cash against perp margin. Verify how aggregate equity/return is computed across the two venue accounts and confirm the headline reconciles with summed trade PnL + funding before re-submitting. (2) THIN EDGE: net carry per pair is frequently swamped by perp-vs-spot exit basis drift plus double-leg commissions (PF 1.03). Confirm realized carry net of basis + fees is robustly positive; consider a basis-aware exit. (3) SAMPLE TOO SMALL: only 14 pairs, all 2024-2025, because WIF/ORDI/SUI spot history starts ~2024. Broaden to alts with longer spot history and/or relax the percentile gate so there are enough pairs to optimize meaningfully. Detailed notes written to workspace/discussions/5cf45434-59ff-40aa-b2d9-9053a7138016/iteration_1_feedback.md.

This delta-neutral spot-perp carry basket set out to capture the funding paid on persistently-premium altcoins, and it proved the plumbing worked — over 6.5 years the engine credited +$22,402 of genuine 8h funding on a truly market-neutral book. But the economics were flat-to-negative (net -1.26%): thin carry barely cleared hedge basis bleed and commissions, and virtually all the profit landed in the 2020-2021 retail-mania window before the top-percentile gate stopped firing. Optimization confirmed the trouble with a failed, negative holdout and an overfit walk-forward fit to a 2020-2021-only sample, and reviewers noted the risk metrics were contaminated by a dual-venue equity-aggregation artifact. The strategy was abandoned after 3 iterations as a target/premise problem — established alts lack the persistent funding premium, and the memecoins that carry it lack backtestable history.

A funding-carry edge is only as durable as the funding premium itself — choosing established alts for backtest history sacrifices the persistent high-retail-demand premium the thesis requires, so the gate stops firing after the mania regime and no parameter sweep can restore a decayed carry source; funding was correctly credited, so the failure is edge decay, not an engine limitation.

It reached optimization, where the holdout failed (ratio -5.615) and the walk-forward was overfit on a 2020-2021-only sample; the backtest-review verdict was 'abandon' and the analyst verdict was 'revise_hypothesis', and the run ended abandoned. The decisive problem was a target/premise failure, not a parameter one: the cohort was swapped to established alts (DOGE/AVAX/LINK/DOT/ADA/SOL/XRP/LTC) for history length, but their positive-funding premium structurally decayed after 2021, while the memecoins that actually carry the edge lacked pre-2024 history.

A delta-neutral positive-funding carry basket that holds matched long-spot (BINANCE_SPOT) plus short-perp (BINANCE USD-M) legs on high-funding altcoins, opening a pair only when an asset's trailing funding sits in the top percentile of its 30-day history and clears an absolute floor, to bank the 8h funding paid by long-perp holders with daily rebalancing and asymmetric hysteresis.

The mechanism worked as designed — the engine credited +$22,402 of real 8h funding cash flow and the book was genuinely delta-neutral (beta ~0) with low fee drag — but the net economic result was flat-to-negative: total return -1.26% with essentially all P&L concentrated in 2020-2021 (~5.97% and ~5.85%), collapsing to near-zero in 2023-2024 and no activity in 2022/2025/2026. Headline risk stats (Sharpe 3.99, and optimized holdout -3.54) were flagged as near-zero-vol / dual-venue-aggregation artifacts and treated as contaminated/unreliable.

EthSpotPerpFundingCarryHysteresisNeutral1H