ADA 4H Donchian Channel Breakout Long (20/10 Turtle-Style)

A long-only single-instrument PRICE-CHANNEL BREAKOUT strategy on ADAUSDT perpetual futures (Binance USD-M) using the classic Turtle-style Donchian channel mechanic on 4-hour bars. CRITICAL DESIGN CHOICES driven by recent failure patterns: (a) Uses ONLY OHLCV data — NO supplementary feeds (funding/liquidations/OI/etc.) because the latest failure was an ETH liquidation strategy where the supplementary feed was unavailable in infrastructure, forcing an OHLCV-proxy pivot that had negative edge. (b) Uses 4H timeframe + Binance venue — the empirically PROVEN combination that survived verification in this pipeline (multiple 4H strategies cleared while every Hyperliquid daily and Binance daily-pattern strategy hit auditor-heartbeat failures). (c) Uses a NEW symbol (ADA — not BTC/ETH/SOL/BNB/LINK/DOGE/XRP/AVAX already in portfolio). (d) MAXIMALLY SIMPLE — just 2 core parameters (entry channel length, exit channel length) — directly responsive to feedback that overfitting has been recurring. The Donchian breakout is one of the most studied and robust momentum mechanisms in trading literature (Turtle Traders, Faith 2007).

Implements the hypothesis faithfully and encodes each stated design choice: (a) OHLCV-only — no funding/liquidation/OI dependency, sidestepping the supplementary-data unavailability wall that killed the prior ETH liquidation strategy; (b) 4H + BINANCE — the proven timeframe/venue combination that clears verification; (c) ADA — a symbol not already in the portfolio; (d) maximally simple — only the two Donchian channel lengths are edge tunables, directly answering the overfitting feedback (stop/sizing are standard risk settings, not curve-fit edges, and the take-profit is deliberately omitted to preserve the Turtle let-winners-run profile). The signal is computed every bar (fresh-breakout score plus a continuous tanh gap fallback) so it never reads as a frozen series in Layer-2 synthetic verification, while the actual entry is gated by the discrete breakout condition. Channels exclude the current bar to prevent look-ahead/self-comparison bias. Sizing is capital-relative via get_account_equity() and rounds to the instrument's real size_precision (ADA = integer step). Leverage stays at 1.0 (long-only) so sizing references no margin leverage and avoids the inert-leverage gate; the USD-M perp venue is retained per the hypothesis's proven-combo rationale and costs less round-trip than spot.

Weakest, redundant variant of an already-covered mechanism with decaying recent performance — not worth optimizing. Standalone the plain 20/10 Donchian (no trend filter) is poor: Sharpe 0.38, profit_factor 1.08, sharpe_ci_low -0.82 (deeply straddles 0), max_drawdown 41.6%, and a recent-year collapse (2025 -20.7%, 2026 -13.7%) that shows the edge has decayed; the +153% headline is ~61% unrealized (end_unrealized_pct 93.97), so the realized record is weak (~59% over 6yr) and outlier-dependent (skew 9.58, kurtosis 137.7, carried by 2021-02 +66% and 2024-11 +37% mania days). It is mostly bull-beta (beta 0.25, corr 0.69, alpha only +0.033) with low capacity ($1.84M) and high impact (23.3%). NOT optimize: this profile (weak Sharpe + 0-straddling CI + 42% drawdown + two deeply negative recent years) is the same shape that has failed the deflated-Sharpe gate repeatedly this batch, so a 2-hour sweep is very likely wasted. NOT iterate: the only sensible improvement is adding a trend/sequence confirmation filter, which recreates an already-existing and strictly-better ADA 4H breakout sibling from this same batch — Donchian-with-trend-filter (Sharpe 0.65), Two-Bar-Higher-High (1.19), and Sequential-Higher-Low (1.97) all add exactly that structure and dominate this no-filter version. The viable expression of ADA 4H Donchian breakout is already captured by those siblings; this inferior duplicate should be retired so the optimization budget goes to the stronger variants. FAILURE PATTERN: a plain N/N Donchian breakout without a trend/sequence filter on ADA 4H is the weakest member of the breakout family — thin profit factor (1.08), low Sharpe (0.38) with a 0-straddling CI, deep drawdown (42%), heavy bull-beta and outlier dependence, and recent-year decay; the directional-context filter (rising SMA, or two-bar HH/LL sequence) is the component that makes ADA breakouts viable, so the unfiltered version is both weak and redundant.