LINK Perp 4H Funding-Extreme Momentum Continuation (Single-Instrument USD-M, Long-Short, Validated-Edge Decorrelated Sleeve)

A long-SHORT, SINGLE-INSTRUMENT, single-venue MOMENTUM-CONTINUATION strategy on LINKUSDT.BINANCE USD-M perpetual futures using 4-HOUR bars, completing a decorrelated multi-asset sleeve of the ONE edge validated this session — funding-LEVEL-Z-extreme momentum-follow (SOL version advanced at PF 1.30). Funding is a SIGNAL only (never credited as a cash flow; P&L is fully price-based and creditable). Thesis: a relative (z-scored) funding EXTREME marks strong, persistent leveraged-positioning demand that CONTINUES in price; the session proved FADING it loses (SOL fade -58%), FOLLOWING the LEVEL extreme works (SOL PF 1.30), the funding DERIVATIVE is anti-predictive (AVAX acceleration PF 0.78), and the settlement-clock variant is fee-dominated (ETH PF 0.88). LINK is chosen deliberately for TWO-SIDED funding: its oracle/DeFi-infrastructure catalyst cycles drive frequent funding SIGN-FLIPS (both extreme-positive and extreme-negative episodes), which keeps the long-short book genuinely two-sided — directly avoiding the funding-sign-asymmetry degeneration (near-one-sided book) that sank the ETH settlement strategy where funding is almost always positive. Its catalyst timing is idiosyncratic relative to the SOL (L1) / DOGE (meme) / XRP (payments) / BTC (major) members of the pending validated-edge book, so it adds a decorrelated sleeve rather than redundant exposure. It stays on the only confirmed-testable substrate (OHLCV + funding, both dense across full history) on the clean single-instrument USD-M MARGIN path, avoiding every structural blocker (NOT supplementary-data-gated/OI/liq/options recent-only, NOT BINANCE_SPOT CASH, NOT COIN-M, NOT cross-venue/HL, NOT multi-instrument pair, NOT funding-carry, NOT breakout, NOT funding-veto overlay, NOT funding-derivative, NOT slow-regime). It is the validated SOL construction (4H funding-Z-extreme, FOLLOW direction) — distinct from the failed funding-fade/acceleration/settlement/regime/veto variants. META-NOTE FOR FACTORY: the OHLCV+funding substrate is now saturated with this edge across 5 assets; unlocking genuinely new mechanism families requires the data layer to backfill multi-year OI/liquidations/options history. Kept to ~4 parameters.

The Layer-3 failure was 'No trades / should_enter returned a side 0 times over 13815 evaluated bars' -- the signal scaffolding ran on every bar but the entry never fired because the funding z-score was never valid. Root cause: _build_funding gated rows on `if not isinstance(k, int): continue`, but the funding_rates dict is documented as dual-keyed {nanosecond_int: rate, "nanosecond_str": rate} and, critically, after the backtest subprocess's JSON round-trip ALL keys arrive as strings (JSON object keys are always strings), so isinstance(k, int) matched nothing -> _fund_ts empty -> _current_funding always None -> _fund_series never filled -> _funding_zscore always None -> _cur_z_valid always False -> no entry. The minimal fix parses every key via int(k) inside a try/except (accepting both int and str keys) and de-duplicates the dual int/str keys into a dict before sorting, so the full funding history loads and the z-score becomes valid once funding_z_lookback samples accrue. This is the same funding-key-stringification trap that silently zeroes funding-driven strategies. Every other line -- the per-bar funding sampling within the 8h settlement window, the level z-score, the FOLLOW entry gate (z >= +/-entry_z), the normalize/flip + stop + max-hold exits, the continuous funding-z+ROC signal that passes Layer-2, and the capital-relative sizing -- is byte-identical to the previous iteration, so Layers 1-2 stay green and the no-price-fallback 'fail loud if funding empty' discipline is preserved. leverage stays 1.0 (used as a non-amplifying sizing base) so no unused-leverage gate applies. This replicates the session-validated SOL funding-Z-extreme momentum-follow construction (PF 1.30) on LINK's two-sided funding for a decorrelated sleeve.

The funding-Z-extreme momentum-follow has no directional edge on LINK either: over a dense, genuinely two-sided 336-trade sample (6.4 years) it loses -12.0% (CAGR -2.6%) with Sharpe -1.38 and the CI entirely below zero (-2.79 to -0.05), PF 0.73, expectancy -$45.8/trade, omega 0.73, and negative in 5 of 7 years. This is a real verdict on the mechanism, not a data artifact (dense sample, balanced long/short, low fee drag). The sleeve's premise is false: it rests on a claimed 'SOL PF 1.30' validation that did not occur — the SOL funding-momentum-follow I reviewed was PF 0.93 (realized-negative) and abandoned. Every asset in this family has now failed under review: BTC (PF 0.65, all-negative Sharpe CI), AVAX funding-acceleration (PF 0.78), ETH settlement-momentum (PF 0.74), DOGE (a 2021-mania kurtosis artifact), and now LINK (PF 0.73). No entry_z/exit_z setting manufactures an edge that is absent across five majors, so optimization would waste 2 hours. FAILURE PATTERN: funding-level/Z direction as a price-continuation signal has no exploitable forward-price edge on any single crypto major — the entire single-instrument funding-as-direction substrate (fade, follow, acceleration, settlement, regime, veto) is exhausted and dead. The hypothesis's own meta-note is right that this OHLCV+funding directional substrate is saturated; the correct conclusion is to STOP proposing single-asset funding-direction strategies and redirect to the one genuinely promising funding construct — the negative-funding-CARRY basket, where funding is the credited cash flow rather than a price-direction predictor — or unblock new mechanism families by backfilling multi-year OI/liquidations/options history.