FMZ funding-rate digest, operationalized: concentrated delta-neutral funding carry on 8 Binance alts (long BINANCE_SPOT / short USD-M perp), top-3 by trailing funding, premium-gated entry, negative-premium exit (1H)

Derived by the Research Lead from 小草, "资金费率策略近况和推荐操作" (FMZ digest, 2021-12-02, https://www.fmz.com/digest-topic/8469, building on the strategy described at https://www.fmz.com/digest-topic/6381). The article describes a hedged perpetual funding arbitrage but does not specify its rules: long spot and short perpetual on coins with high positive funding, opened when the perp trades at a premium to spot (开仓溢价) and closed when that premium turns negative (平仓溢价). It makes three recommendations: (1) concentrate in a FEW high-funding coins rather than spreading across many, (2) use more exchanges, (3) seize extreme-funding events. What the figures show: two position tables from the author's live robot (digests 8469 and 6381) list per coin the short perp quantity, the matching spot quantity, an open premium % (typically 0.03–0.3%), a close premium % (0.1–0.18%), target position, average funding % (about 0.03%/8h in Dec-2021, versus about 0.15–0.17%/8h earlier in the bull market), current funding and reference annualised yield. A cross-exchange funding table shows OMG at -2.4% (Binance) to -5.4% (FTX) during an airdrop. Two further images are banners with no data. The linked code (strategy 333315) is only a funding-rate dashboard across exchanges. The robot page could not be read (FAILED). No exact rules were available, so the rules below are mine. OPERATIONALIZATION: the universe is 8 alts that appear in the author's own position tables and are listed on both Binance Spot and Binance USD-M: SAND, ICP, VET, ATOM, ANKR, 1INCH, EGLD, IOTA. Each coin is a pair of legs of equal USDT notional: long <COIN>USDT.BINANCE_SPOT (cash) and short <COIN>USDT.BINANCE (perp). The book is therefore delta-neutral, and PnL is funding received plus basis change minus fees. All 16 instruments use 1-HOUR bars. Funding is read point-in-time with supp_as_of / supp_window on the perp, never by abs() timestamp matching. At each 8h funding settlement (00/08/16 UTC bar close), coins are ranked by trailing mean funding over the last `funding_lookback` settlements. Only the top `top_n` whose mean exceeds `entry_funding_min` are held (recommendation 1, concentration). Recommendation 2 (multiple exchanges) is deliberately left out of this test: keeping one exchange means both legs share a venue's funding and prices. Recommendation 3 (OMG-style extreme negative funding) would need borrowed spot to short, which a CASH account cannot do, so only the positive-funding side is tested. Execution: every order is a market order at the signal bar's close, which is the 24/7 equivalent of a next-bar-open fill. The spot leg is long-only and unleveraged; the perp short is at 1x notional matching the spot leg, with leverage fixed at 1.0. Cost arithmetic: a full pair round trip pays 2×0.10% spot + 2×0.05% perp = 0.30% of one leg's notional (0.15% of gross two-leg notional). The source's late-2021 average funding is about 0.03%/8h (0.09%/day). Over a minimum hold of 7 days (21 settlements), expected funding is about 0.63% per pair round trip before basis. Adding the entry premium of at least 0.05% as basis converges gives about 0.68% gross versus 0.30% cost per leg-notional, so about 0.38% net per pair, or about 0.19% of gross notional per leg-trade. In the higher-funding regimes the source shows (0.15%/8h), the capture is several times that. This is a cross-venue (BINANCE_SPOT + BINANCE), market-neutral, multi-instrument hypothesis. Those buckets are under-represented in the portfolio (cross-venue 7.9% versus the ≥15% target).

Implements the FMZ 8469/6381 hedged funding arbitrage as operationalized. Recommendation 1 (concentration) is the top_n trailing-funding ranking. The source's open premium and close premium are open_premium and the negative close_premium, with premium computed from contemporaneous closes of the perp and spot bars (the base sync barrier aligns all 16 legs; a leg bar newer than the decision bar is never used). The funding_flip, rank and hedge-break exits plus the minimum hold follow the plan's fixed and tunable parameters, with exact names. Funding is looked up by bisect_right at the bar timestamp (point in time, no abs() matching). Records are normalised to per-8h because some Binance alts moved to 4h funding intervals; without that, rankings would mix units. A coin without funding data is simply never eligible; there is no price proxy. The ranking is a pure function of the clock and the hold time comes from the perp position's ts_opened, so a restart changes no decision. Both legs use an identical coin quantity, floored to the coarser size increment, so the book is delta-neutral. Spot and perp sit on separate $100K venue accounts, each at 0.95x of its own equity (leverage 1.0, nothing to wire). Every leg is its own order and position, so total_trades counts legs (lesson 146). Expect roughly 2 x top_n legs per rotation. Fee note: a pair round trip costs 0.30% of one leg's notional. The minimum hold of 21 settlements and the 0.02%/8h entry floor target about 0.4%+ funding per hold, but in 2023-2025 alt funding sits near the 0.01% base, so few coins may clear entry_funding_min outside bull phases.