BtcFourHourFailedBreakdownReclaimMultiBarLong

BTC 4H Failed-Breakdown Reclaim Long (Multi-Bar Structural Stop-Hunt Pattern)

A long-only, single-instrument strategy on BTCUSDT.BINANCE 4H bars implementing a MULTI-BAR STRUCTURAL pattern — the analyst's explicit recommendation from the ADAHammerReclaimAbsorptionLong abandon feedback ('MULTI-BAR PATTERN ENSEMBLES — at least adds confirmation that single-bar shapes lack'). The mechanism: identify the 10-bar Donchian low (excluding the current bar) as a structural reference level; detect a BREAKDOWN event when any of the last 2 bars made a new low BELOW that reference; detect a RECLAIM event when the current bar CLOSES BACK ABOVE the reference level AND closes bullishly (close > open). The combination (breakdown + reclaim within 2 bars + bullish close) signals a 'failed breakdown' / stop-hunt absorption — algorithmic flow swept liquidity at the obvious low, the sweep failed to trigger sustained selling, and buyers absorbed and reclaimed the level. Critically distinct from the failed ADAHammerReclaimAbsorptionLong (PF 0.45 abandon) because: (a) requires an ACTUAL structural breakdown of a historical price level, not just a single-bar lower wick (a wick may not be a real breakdown if it doesn't violate the 10-bar reference); (b) the reclaim is tied to a specific REFERENCE PRICE LEVEL (the prior 10-bar low), not just bar-body geometry; (c) MULTI-BAR pattern (breakdown bar at t-1 or t-2, reclaim bar at t) vs single-bar geometry; (d) applied to BTCUSDT (institutional liquidity, less retail noise than ADA). Trend-filtered (close > sma_100) to prevent fading actual downtrends per the Keltner-failure lesson. Long-only because (i) the absorption thesis works most reliably in confirmed uptrends, (ii) short-only directional on crypto majors has been empirically refuted. Single-instrument on confirmed-working BTC 4H bars (BtcFourHourVolumeBreakoutLong validates this bar type).

Implements the analyst's explicit multi-bar-ensemble prescription and fixes every flaw of the abandoned ADAHammerReclaimAbsorptionLong (PF 0.45): (a) it requires an ACTUAL breakdown of a historical reference PRICE LEVEL — the Donchian low computed over a window that excludes the current bar and the breakdown bars, so the sweep is measured against a level set before it — not a single-bar wick; (b) the reclaim is defined relative to that reference level, not bar-body geometry; (c) it is genuinely multi-bar (breakdown at t-1/t-2, reclaim at t); (d) it runs on BTC's institutional liquidity. The SMA(100) uptrend filter prevents fading real downtrends (the Keltner lesson), and long-only matches the absorption thesis's reliability in uptrends and the empirical refutation of short-only crypto-major directional bets. The swept low doubles as the structural stop, giving a tight, well-defined risk that drives risk-based sizing (avg_win can exceed avg_loss). calculate_signal returns a continuous distance-above-reference each bar (clamped to ±50) so Layer-2 synthetic scenarios are never frozen and no invalid (>1e9) signal can occur. Leverage stays at 1.0 so sizing references no margin leverage (avoiding the inert-leverage gate); the bar type BTCUSDT.BINANCE-4-HOUR is the validated, reliable data path.

Near-zero edge on an adequate sample — a genuine no-edge result, not optimizable or under-sampled. With 176 trades (~29/yr, well above the walk-forward floor), the strategy returns Sharpe 0.28, profit_factor 1.08 (barely above breakeven), alpha +0.0007 (~zero), CAGR 1.69%, and sharpe_ci_low -1.02 (deeply straddles 0). The entire +21.9% total return comes from a single year (2024 +21.9%); the other six years are flat-to-negative (4 of 7 negative: 2020 -0.7%, 2021 -0.6%, 2022 -7.0%, 2023 -2.2%, 2025 -1.3%), so the pattern only 'worked' in one strong-trend regime. NOT optimize: PF 1.08 / Sharpe 0.28 will fail the deflated-Sharpe gate, and a 225-trial sweep would curve-fit to the 2024-specific configuration. NOT iterate: this is the key distinction from the thin-sample candidates I sent back for density (BNB, BTC/ETH-TSM, ETH volume-surge) — those had ~35-42 trades, but THIS has a full 176-trade sample that already demonstrates the edge is absent (alpha ~0), so it is not a density/parameter problem and tuning cannot manufacture predictive power that 176 trades show isn't there. The failed-breakdown-reclaim / absorption pattern family is empirically weak on majors: the prior single-bar ADAHammerReclaimAbsorption was abandoned at PF 0.45, and this stronger multi-bar structural version only reaches PF 1.08 — still no real edge. FAILURE PATTERN: structural failed-breakdown reclaim (Donchian-low sweep + bullish reclaim + trend filter) on BTC 4H generates an adequate trade count but no predictive edge — PF ~1.08, alpha ~0, Sharpe 0.28 with the entire return concentrated in one trend year (2024) and 4 of 7 years negative; the reclaim/absorption pattern family does not predict continuation on crypto majors, single-bar or multi-bar.

Long-only, single-instrument BTCUSDT.BINANCE 4H multi-bar failed-breakdown reclaim. It takes the Donchian low over the 10 bars ending BEFORE the last 2 bars as a structural reference, requires one of the last 2 bars to have swept a new low below that reference (breakdown), then enters LONG when the current bar closes back above the reference with a bullish body (reclaim) while price is above SMA(100). Exits on a close back below the swept low (structural invalidation), an 8% take-profit, or a 30-bar time cap. Positions are risk-sized (1% of equity to the structural stop, capped at 50% notional). Pure OHLCV, leverage 1.0.

Clean execution, long-only as designed (176/176 submitted, no artifacts), adequate sample (176 trades), low drawdown (14%); sound multi-bar structural construction (genuine reference-level breakdown+reclaim, not single-bar geometry) with a real structural stop

Distinct from and better than the abandoned single-bar ADAHammerReclaim (PF 0.45 -> 1.08)

Near-zero edge: Sharpe 0.28, profit_factor 1.08, alpha +0.0007 (~zero), CAGR 1.69%, sharpe_ci_low -1.02 (deeply straddles 0)

Entirely single-year-dependent: only 2024 positive (+21.9% = the whole total return); 4 of 7 years negative (2020/2021/2022/2023/2025)

Adequate 176-trade sample shows the edge is genuinely absent (not a density/thin-sample issue); fee drag (commission 5.29% of gross) on a razor-thin gross edge

Reclaim/absorption pattern family empirically weak on majors (prior ADAHammerReclaim abandoned)

meaningful > 0

adequate (so no-edge, not under-sampled)

only 2024 positive; 4/7 years negative

majority positive

BtcFourHourFailedBreakdownReclaimMultiBarLong was a deliberate, analyst-prescribed improvement on the abandoned single-bar ADAHammerReclaim: a genuinely multi-bar structural pattern requiring an actual sweep of a 10-bar Donchian reference low followed by a bullish reclaim, trend-filtered and run on institutional-liquidity BTC. The construction was sound and clean — and it did beat its predecessor (PF 0.45 → 1.08) — but it still produced no real edge: Sharpe 0.28, alpha ~0, sharpe_ci_low straddling zero, with the entire +21.9% return concentrated in 2024 and four of seven years negative. The reviewer abandoned it pre-optimization, emphasizing that 176 trades is an adequate sample, so the absent edge is real rather than under-sampled — the decisive distinction from the thin-sample candidates sent back for density. The conclusion was that the reclaim/absorption pattern family simply does not predict continuation on crypto majors, single-bar or multi-bar, so there is nothing to tune toward.

The failed-breakdown reclaim / absorption pattern family does not predict continuation on crypto majors in either single-bar (ADAHammerReclaim, PF 0.45) or stronger multi-bar structural form (PF 1.08, Sharpe 0.28) — and when an adequate trade count (176) shows alpha ~0 with the return concentrated in one regime, it is a true absent-edge result, not a density or parameter problem to iterate on.

The analyst abandoned it at the backtest-review gate, before optimization, as a genuine no-edge result rather than a thin-sample one: 176 trades (~29/yr, well above the walk-forward floor) already demonstrate the edge is absent (alpha ~0), the whole return came from a single 2024 trend regime, and PF 1.08 / Sharpe 0.28 would fail the deflated-Sharpe gate — so optimization would only curve-fit the 2024 configuration and tuning cannot manufacture predictive power that isn't there.

A long-only, single-instrument BTCUSDT.BINANCE 4H multi-bar structural pattern: detect a breakdown when recent bars sweep below a 10-bar Donchian reference low, then enter long when the current bar reclaims that level with a bullish close (a 'failed breakdown'/stop-hunt absorption), gated by a 100-period SMA uptrend filter with a structural stop at the swept low — the analyst's prescribed multi-bar upgrade of the abandoned single-bar ADAHammerReclaim pattern.

Over an adequate 176-trade sample it had near-zero edge: +21.9% total return but CAGR just 1.69%, Sharpe 0.28, profit factor 1.08, alpha +0.0007, sharpe_ci_low -1.02, 40.3% win rate, 14% max drawdown — with the entire total return concentrated in 2024 and 4 of 7 years negative; execution was clean (176/176 submitted, metrics_reliable=true, beta 0.066).