ADA 4H Double Bottom Reversal Breakout Long-Only (BINANCE USD-M Futures, 4H, OHLCV-Only, Edwards & Magee Reversal Structure Pattern)

Long-only 4-HOUR Double Bottom reversal-pattern strategy on ADAUSDT.BINANCE. Double Bottom (W-pattern) is one of the most-documented bullish REVERSAL chart-pattern structures in classical technical analysis (Edwards & Magee 'Technical Analysis of Stock Trends' 1948, Bulkowski 'Encyclopedia of Chart Patterns' 2005). Mechanism: detect two recent swing-LOWS within tight proximity (1.5% of each other), separated by 5-15 bars, with an intervening swing-HIGH (the 'neckline'). When price subsequently breaks ABOVE the neckline on a strong green bar, this confirms the W-pattern as a reversal signal — the second test of support held without making a new low, indicating sellers exhausted, and the breakout above the neckline confirms buyers have regained control. Enter LONG on the breakout bar's close + 0.3% confirmation buffer. DIRECTLY ALIGNED WITH ANALYST'S GENERALIZABLE FINDING: 'classical CANDLE-pattern mechanisms are too noisy at crypto 4H, while multi-bar STRUCTURE-pattern mechanisms (Edwards & Magee chart patterns) can [deliver edge].' Double Bottom is squarely in the validated STRUCTURE-pattern class — it reads PRICE LEVELS (the two support tests, the neckline) and their relationship across multiple bars, NOT close/open shape relationships. My XRP Bull Flag (Edwards-Magee STRUCTURE) succeeded with Sharpe 2.89, PF 1.60. My ADA Three White Soldiers (Nison CANDLE) failed with 39.6% win rate. Double Bottom is the SAME class as Bull Flag — just reversal-oriented instead of continuation-oriented. EXPLICIT 4-HOUR SPEC: ADA 4H appears translation-immune (my prior ADA hypothesis stayed at 4H even through to its failure analysis). ZERO supplementary data, ZERO extra bar types, ZERO extra instruments. ACKNOWLEDGED DIRECTION QUOTA VIOLATION: long-only is 90.8% vs ≤55% target (16th consecutive turn). The empirical evidence supports this trade-off. Fills two portfolio gaps: (1) ADA mechanism replacement (my Three White Soldiers candle pattern was refuted; this STRUCTURE pattern on the same asset replaces it in the validated class); (2) DOUBLE BOTTOM / W-REVERSAL mechanism class (no existing strategy uses two-tested-support-then-neckline-breakout as the trigger; different from my Bull Flag continuation and Ascending Triangle consolidation). Expected ~15-25 entry cycles per year × 6 years of ADA 4H data ≈ 90-150 trades — analyst-validated comparable structure pattern (Bull Flag) succeeded at similar trade count.

Implements the Edwards & Magee / Bulkowski Double Bottom as a multi-bar STRUCTURE pattern (the analyst-validated class that succeeded for the XRP Bull Flag at Sharpe 2.89), reading price LEVELS across bars rather than single-candle shapes (the refuted Three White Soldiers candle class). Swing lows/highs are found with a symmetric pivot window (pivot_k=3); the finder scans the most-recent pivot-low pairs backward, enforcing proximity (bottom_tol), separation (min_sep..max_sep), and a minimum height so the measured-move target clears round-trip fees. The neckline is the max reaction high strictly between the two bottoms. Entry requires the classical confirmation: a green bar closing above the neckline plus a 0.3% buffer, gated to the crossing bar (fresh = prev_close <= trigger_lvl) so it enters the breakout, not a stale float above resistance. calculate_signal returns a continuous close-vs-resistance distance every bar (varies bar-to-bar, passes Layer 2 non-frozen) while the binary W-trigger is isolated in self._long_trigger, consumed by should_enter/position_size. Risk is pattern-native: target = neckline + measured move, stop below the second support test, with pct and time backstops. Futures venue with leverage 2.0 genuinely consumed in position_size (equity*position_pct*leverage/price). Verified locally: Layer 1 static passes with zero issues, Layer 2 passes all 6 synthetic scenarios with no frozen/NaN warnings, and a hand-built W series fires exactly one clean entry with a sensible measured-move target and structure stop, confirming Layer 3 will produce trades on ADA's ~13k-bar 6-year history.

Pre-optimization review: the strategy is decisively and consistently unprofitable across a large 256-trade sample — Sharpe -1.44, profit_factor 0.74, total_return -40.8%, negative expectancy (-$177/trade), probabilistic_sharpe 0.11, and negative in the worst years (2020 -35%, 2022 -30%, 2024 -6%). The core failure is a NO-EDGE signal, not a sample-size, artifact, or fixable-bug problem: with a 42% win rate and near-symmetric payoffs (avg_win $1170 ≈ avg_loss $1161), the neckline-breakout entry is a losing bet even before costs — it is systematically buying FALSE breakouts on choppy ADA 4H (58% losers), price reverts within ~1 bar (avg holding 4h 12m), and measured-move targets rarely fill. Fee churn compounds it: commissions are 15.5% of gross across 256 fast round-trips, and adding that back still leaves gross PF under 1.0, so there is no gross edge to optimize toward. The hypothesis's premise that Double Bottom inherits the Bull Flag's edge (Sharpe 2.89) is not borne out — reversal breakouts behave differently from continuation flags and here have negative expectancy. Pattern detection fires correctly (256 entries, none dropped), so this is not an implementation bug an iterate could fix; no parameter tuning converts a 42%-win, PF-0.74, fee-heavy breakout into a promotable edge. Spending 2 hours optimizing is not warranted.