COIN-M BTC Trend-Follower (Dual-EMA Crossover + Chandelier Trailing Stop) — Run the ONE Proven Edge on the LOWEST-FEE, Under-Represented BINANCE_CM Inverse Venue Where the Fee Hurdle Is 40% Smaller (BTCUSD_PERP.BINANCE_CM, Daily Bars, Long-Short, Asymmetric Let-Winners-Run, 3-Parameter)

A LONG-SHORT, single-instrument, ASYMMETRIC-PAYOFF trend-follower on BTCUSD_PERP.BINANCE_CM — the CURRENT coin-margined (inverse) BTC perpetual on Binance COIN-M (explicitly the perpetual, NOT a dated/expired quarterly, per L58). It rides the ONLY edge the factory has ever promoted (directional BTC trend with a trailing stop that lets winners run) but makes two deliberate diversification moves versus the five pending single-asset trend sleeves (all on BINANCE USD-M or Hyperliquid): (1) VENUE — COIN-M is under-represented (3.4% vs a ≥5% target) and, decisively, carries the LOWEST futures fee in the factory (~0.06% round-trip taker vs ~0.10% on USD-M). Since the #1 killer across 1164 experiments is fee_edge — dozens of otherwise-real edges died grazing the 0.15% USD-M floor — running trend on the venue where the round-trip hurdle is 40% smaller is a direct, quantified attack on that failure mode. (2) CONSTRUCTION — a dual-EMA CROSSOVER regime with a chandelier trailing exit, mechanically distinct from the pending Donchian-breakout / momentum-confluence / MA-side / OI-gated constructions (so it is not L56 symbol-spray of one template). Pure OHLCV, computationally TRIVIAL (two EMAs + ATR, a few scalars per bar, no full-history rescans, no sorted()/min() in the hot path). Avoids every graveyard: NOT symmetric reversion/pairs (all died fee_edge/no_edge this session), NOT funding carry/differential (L57), NOT cross-sectional (died overfit), NOT options (infra-broken), NOT a liquidation feed (L58). Exactly 3 tunable parameters: fast EMA span, slow EMA span, and chandelier ATR multiple.

Verification failed (Layer 4 — QA review) [class=no_edge]: - UNMEASURABLE + NO ALPHA at the specified single-instrument daily frequency. 31 trades over 5.8 years: the hard holdout gate needs >=10 trades in the last-20% holdout, but ~6 land there, so it cannot pass by construction and optimization dies there (same trade-starvation as the BTC CM daily Donchian [36], ETH CM Donchian [26] and CM dual-EMA siblings). It is BTC beta, not skill: alpha -0.0116 (negative), IR -0.74, beta 0.19, benchmark_correlation 0.82 to buy_hold — the +48.6% is partly an open-position markup (end_unrealized +27.6%). Sharpe 0.471 with CI [-0.16, 1.14] straddles zero (below the 0.5 OOS floor), kurtosis 8.9 (PnL in a few legs). avg_trade 4.99% and PF 1.74 are a tiny, tail-driven sample's artifact, not a robust edge — and fees were never binding (avg_trade clears the CM floor ~50x), so the low-fee-venue thesis doesn't rescue it. - The developer's dry run (32 trades, 39% win, +8% avg gross) matches the sandbox's tail-driven few-large-winners profile, and the developer flagged the modest-sample tail-sensitivity. Construction is sound (no liquidation, correct sign, lockout prevents churn), but a directional-OHLCV daily trend book's base Sharpe is its ceiling, and here that base is a straddling-zero 0.471 with negative alpha.