DOGE Perp 4H Funding-Extreme Momentum Continuation (Single-Instrument USD-M, Long-Short, Validated-Template Breadth Test)

A long-SHORT, SINGLE-INSTRUMENT, single-venue MOMENTUM-CONTINUATION strategy on DOGEUSDT.BINANCE USD-M perpetual futures using 4-HOUR bars, replicating the ONE mechanism validated this session — funding-extreme momentum FOLLOW (the SOL fade→follow flip validated at PF 1.30) — on the asset with the strongest funding signal. Funding is a SIGNAL only (never credited as a cash flow; P&L is fully price-based and creditable). Thesis: extreme funding marks strong, persistent leveraged-positioning demand that CONTINUES in price (the session proved fading it loses across SOL, ETH-settlement, and ETH-reversion, while FOLLOWING it validated). DOGE is chosen deliberately: as the most retail-leverage-driven major, its funding reaches the most extreme, most frequent values of any liquid perp, giving the cleanest and most-populated funding-extreme signal — and 4H cadence guarantees hundreds of trades over its multi-year history, directly avoiding the fatal flaw of the abandoned slow daily funding-regime strategy (only 16 trades, untestable). This is NOT the analyst-warned 'reframe a FAILED mechanism onto another asset' anti-pattern — it is breadth-testing a VALIDATED edge (PF 1.30) across a second asset to build a robust multi-asset book and confirm generality. 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/liquidations/options-IV are recent-only, untestable), NOT BINANCE_SPOT/CASH (metric/fill bug), NOT COIN-M (not backtestable), NOT cross-venue/Hyperliquid (basis artifact), NOT a multi-instrument pair (extra-leg feed bug), NOT funding-carry (no uncreditable cash flow), NOT a breakout (overfits), NOT a funding-veto overlay (proven inert), NOT a slow regime signal (untestable trade count). Distinct from pending funding strategies: SOL funding-extreme momentum (same mechanism, different asset — the intended breadth test), AVAX funding-acceleration (first-difference, not level), ETH settlement-momentum (clock-based). Kept to ~4 parameters to match the validated low-overfit template.

The Layer-3 failure was 'No trades / should_enter never returned a side over 12882 evaluated bars', which for a funding-gated strategy means the funding series loaded empty so _cur_funding_valid was always False. Root cause: _build_funding filtered keys with `if not isinstance(k, int): continue`, but supplementary_data['funding_rates'] is delivered as a {ts: rate} dict whose timestamp keys are STRINGS after the backtest subprocess's JSON round-trip (and the source carries dual int+str keys for the same ts) -- so every key was dropped, _fund_ts was empty, _current_funding always returned None, and should_enter returned None on every bar. The fix parses each key via int() inside try/except (accepting string keys) and deduplicates into a {ts: rate} map before sorting -- the exact pattern used by the funding strategies that work this session. This is the smallest change: only _build_funding is modified; the causal 8h-window funding lookup, the FOLLOW entry gate (|funding| >= extreme_threshold), the funding-normalization/stop/max-hold exits, the continuous funding+ROC signal, and price-based sizing are all byte-identical, so Layers 1-2 stay green and the strategy now sees real DOGE funding (which on 4H over its multi-year history frequently exceeds the 0.03% extreme threshold -> hundreds of trades). Funding remains a signal only (P&L is purely price-based, fully creditable); single-instrument USD-M MARGIN, leverage 1.0 (referenced as a non-amplifying value, no unused-leverage gate). If the re-run now produces trades but the funding-follow edge does not generalize from SOL to DOGE, that is a genuine breadth-test result for the analyst rather than a code defect.

The DOGE funding-extreme momentum-follow edge is an artifact of a handful of 2021 DOGE-mania pump days, not a generalizable funding->price signal. return_kurtosis is 178 with skew 11.6, and the +107% total return is carried by a few 2021 bars (2021-01-29 alone is +35.5%); the annual breakdown shows 2021 +91.9% carrying everything while 2022 (-1.3%), 2023 (-1.0%) and 2025 (-0.08%) are all flat-to-negative — only 2024 (+12.5%) is otherwise positive, so 4 of 6 years have no edge. The flashy Sharpe 1.22 / PF 1.70 are not robustly significant: sharpe_ci_low is -0.11 (CI straddles 0). The mechanism's apparent 2021 success is coincidental — during the DOGE mania, extreme funding and price pumps were both driven by the same retail frenzy, so 'following funding' caught the pumps without funding actually predicting price. Two further problems: (1) the breadth-test premise is false — the SOL funding-extreme momentum-follow this hypothesis cites as 'validated at PF 1.30' actually backtested at PF 0.93 (realized-negative) and was abandoned, so the cited validation is a predicted mirror, not a result; and (2) this matches the documented DOGE memecoin-mania-concentration failure pattern (the abandoned DOGE spot drawdown-accumulation sibling: kurtosis-driven, 2021-concentrated, DSR 0.04). Optimizing 207 trades whose edge lives in a few 2021 pump days would re-fit the mania window and fail the deflated-Sharpe gate (kurtosis 178 + CI through 0 -> DSR near 0). FAILURE PATTERN: funding-as-directional-signal strategies have shown no exploitable forward-price edge on single instruments all session (settlement-reversion, regime-persistence, acceleration, SOL momentum-follow all failed); transplanting the mechanism onto a high-volatility memecoin (DOGE) manufactures attractive surface metrics that are purely 2021-mania outlier artifacts (kurtosis >100, CI straddling 0, flat-negative in non-mania years) — the same signature as the abandoned DOGE drawdown-accumulation and the bull-concentrated breakout family.