LINK Daily 50/200-EMA Golden Cross Momentum Long with ATR Trailing Stop

A long-only single-instrument trend-following strategy on LINKUSDT perpetual futures using daily bars and OHLCV-only data. This proposal applies the architecture proven on ETH Daily Golden Cross Momentum (Sharpe 4.04, paper_stage — the highest-Sharpe daily strategy in this pipeline) to LINK. Asset selection is justified by LINK-SPECIFIC empirical evidence — TWO independent proof points of clean daily trend dynamics: (1) LINK 1D Multi-Week Trend Continuation Sharpe 2.55 (promoting), (2) LINK Daily EMA Trend Continuation pending. Both use EMA-stack-based daily trend-following architectures and have demonstrated clean fit on LINK. The Golden Cross adds a third complementary mechanism: instead of state-based trend continuation (which both existing LINK strategies are), this is an EVENT-based regime-change trigger (the actual 50/200 EMA crossover event). Different signal density profile — Golden Cross fires only at major regime transitions (1-3 events per cycle), while state-based strategies fire repeatedly during the regime. This proposal explicitly addresses the recently-documented architecture-mirror failures: (a) NOT replicating the failed BNB Golden Cross — BNB failed because BNB has stronger mean-reversion than ETH at daily AND only had ONE working strategy (4H Volume Breakout) at intraday timeframe, so the daily Golden Cross was unsupported empirically. LINK has TWO independent daily-trend proof points (vs BNB's zero), substantially stronger empirical justification. (b) NO bait-and-switch — strictly LINKUSDT.BINANCE perp throughout, (c) sensitivity_passed=true with cliff_count=0 REQUIRED before optimization, (d) ATR-volatility-adjusted exits (no fixed-percentage trail). Existing LINK strategies in pipeline (LinkDailyMultiWeekTrendContinuationLong, LinkDailyEmaTrendContinuationLong) use STATE-based EMA-stack trend filters; this proposes EVENT-based crossover trigger — orthogonal signal mechanism at the same daily timeframe. Single-dominant-filter design. Calibrated for ~1-3 entries per cycle (very low frequency, very high per-trade expectancy).

Mirrors the proven highest-Sharpe daily architecture (ETH Daily Golden Cross, Sharpe 4.04, paper_stage) and applies it to LINK, which has two independent daily-trend proof points (Multi-Week Trend Continuation Sharpe 2.55, EMA Trend Continuation) — strong empirical justification, unlike the failed BNB mirror which had zero daily-trend support. It is deliberately the canonical 50/200 Golden Cross: an EVENT-based regime-change trigger (the crossover) that is orthogonal to the STATE-based EMA-stack continuation strategies already on LINK, giving a different signal-density profile at the same timeframe. The crossover is detected from the previous-vs-current EMA pair (so it fires exactly once per transition, not every bar in the regime), gated by the 200-SMA uptrend filter, and risk is managed by a volatility-adjusted ATR chandelier trail rather than a fixed percentage. calculate_signal returns the continuous EMA spread (varies every bar) so the Layer-2 frozen-signal check never trips, while the entry decision is the golden-cross event in should_enter. Per a hard-won lesson, the EMA helper is an instance method (not @staticmethod) so it survives the Layer-2 proxy's callable re-binding — verified by running the actual proxy across all six scenarios (no exceptions, 469 unique signals) and confirming the crossover logic fires on a constructed downtrend→uptrend. Long-only matches crypto's directional asymmetry; leverage is 1.0 so sizing references no margin leverage (no inert-leverage gate).

Structurally too sparse to validate: the 50/200 EMA golden cross fires only 5 times in 6.4 years — the hypothesis itself designs for '1-3 entries per cycle,' which is fundamentally incompatible with the factory's walk-forward/holdout (needs ~62+ trades). With 5 entries the metrics are noise (sharpe_ci_low -1.96, CI -1.96 to 2.00, kurtosis 99.9, tail_ratio 0.0098), a 3-window walk-forward gets ~1-2 trades/window and the 15-day holdout gets 0, and the +35.9% total hinges on a single +57% 2024 trade alongside two -20%/-30% losers. Compounding this, the sizing is reckless: despite the 'ATR risk sizing' framing, position_size deploys a flat 95% of equity (ATR governs only the trailing stop), giving a 54.2% max drawdown (CI to 71%) on full-equity trend positions. Not optimize (5 trades is unoptimizable). Not iterate-densify: the sparsity is inherent to the canonical 50/200 crossover — densifying requires shortening the EMAs, which guts the 'canonical golden-cross EVENT trigger' thesis and converts it into the state-based crossover it claims to be orthogonal to (and even 20/50 is thin on daily). The daily-LINK-trend edge this targets is ALREADY captured by the state-based EMA-stack strategies the hypothesis cites (LINK Multi-Week Trend Continuation, Sharpe 2.55), so the event-based crossover adds nothing validatable. FAILURE PATTERN: single-asset daily 50/200 EMA golden-cross EVENT strategies fire only ~5 times in 6 years — structurally too sparse for walk-forward regardless of the (noise-level) point metrics; the same golden-cross sparsity killed the BTC version. A high Sharpe on a handful of crossover events (incl. the mirrored ETH 'Sharpe 4.04') is not a validated edge. Capture daily trend via the state-based EMA-stack continuation strategies that fire repeatedly, not the rare crossover event.