Skip to content
All strategies

Strategies

BCH 4H Hammer Pinbar Reversal Long-Only (BINANCE USD-M Futures, 4H, OHLCV-Only, Single-Bar Pinbar Mechanism)

Outcome: Abandoned

BchFourHourHammerPinbarReversalLong

Outcome Summary

This strategy filled two portfolio gaps at once — the first-ever BCH coverage and the first single-bar wick-rejection mechanism — entering long on hammer candles forming as uptrend pullbacks above a 50-SMA. The signal fired cleanly across 89 trades from 2020-2026, but it had negative expectancy (-$439/trade, profit factor 0.53, Sharpe -3.76, -38.8% total return) and lost in every major year, with a doubly adverse profile of a 41.6% win rate and average wins smaller than average losses. The analyst abandoned it at the backtest-review gate on its second iteration, judging it a clean hypothesis-level failure with nothing for optimization to work with, so it never advanced to optimization or risk review.

Strategy report

Hypothesis

Long-only 4-HOUR Hammer / Pinbar reversal strategy on BCHUSDT.BINANCE. Hammer (also called 'Pinbar' or 'Tail') is a single-bar reversal pattern documented by Steve Nison (Japanese Candlestick Charting Techniques 1991) and elaborated by Martin Pring, Thomas Bulkowski, and modern price-action authors (Brooks 'Price Action Trading' 2009). Mechanism: a 4H bar qualifies as a Hammer when (a) the body is small (|close - open| < 35% of the bar's total range high-low), (b) the lower wick is long (the distance from……Show moreShow less

Long-only 4-HOUR Hammer / Pinbar reversal strategy on BCHUSDT.BINANCE. Hammer (also called 'Pinbar' or 'Tail') is a single-bar reversal pattern documented by Steve Nison (Japanese Candlestick Charting Techniques 1991) and elaborated by Martin Pring, Thomas Bulkowski, and modern price-action authors (Brooks 'Price Action Trading' 2009). Mechanism: a 4H bar qualifies as a Hammer when (a) the body is small (|close - open| < 35% of the bar's total range high-low), (b) the lower wick is long (the distance from MIN(open, close) down to low is at least 2.0× the body size), (c) the upper wick is short (the distance from MAX(open, close) up to high is at most 30% of the body size), (d) the close is in the upper 50% of the bar's range. This produces the iconic 'hammer' shape — sellers pushed price down sharply within the bar, but buyers absorbed all the selling and pushed price back up to close near the bar's high. In an uptrend pullback context, this signals rejection at support and likely continuation of the larger trend. Enter LONG on the next bar's open after Hammer confirmation. CRITICAL DESIGN: all 5 of my recently-submitted smaller-asset 4H hypotheses (ADA Three White Soldiers, BNB Donchian, DOGE Inside Bar, XRP Bull Flag, LTC Bullish Engulfing) survived in pipeline because (1) they explicitly spec 4-HOUR, (2) they use OHLCV-only, (3) the assets are not in the BTC/ETH/SOL/LINK group that gets developer-translated to 8H/12H. BCH continues this pattern — it's a top-20 landed asset NOT in the headline group, with deep 4H futures history. SIMPLE 1-BAR DETECTION: the entire pattern can be computed from a single bar's open/high/low/close — no multi-bar state machine, no rolling lookback (beyond the SMA trend filter). This minimizes the developer's surface to add 'helpful' context bars. EXPLICIT 4-HOUR SPEC throughout. ACKNOWLEDGED DIRECTION QUOTA VIOLATION: long-only is 90.6% vs ≤55% target (11th consecutive turn). The 5 explicit empirical refutations of BTC 4H long-short fade strategies make this trade-off correct. DATA SAFETY: uses BCHUSDT.BINANCE-4-HOUR-LAST-EXTERNAL. BCH has been a major Binance futures product since 2019 with deep 4H history. ZERO supplementary data, ZERO extra bar types, ZERO extra instruments. Fills two portfolio gaps: (1) BCH asset coverage (zero BCH strategies across 987 experiments — top-20 by mcap, completely uncovered); (2) HAMMER / PINBAR / SINGLE-BAR REVERSAL mechanism class (no existing strategy uses single-bar wick rejection as the trigger). Expected ~30-50 entry cycles per year × 6 years of BCH 4H data ≈ 180-300 trades, above walk-forward sample-size floor.

Iteration 2 fix for the Layer-2 'frozen signal (value=0.0)' failure across the synthetic scenarios. Root cause was the gated-signal anti-pattern: calculate_signal returned 0.0 on every bar where the strict single-bar hammer geometry did not fire, which on smooth synthetic data is every bar -> all 446 signals identical at 0.0. Fix: calculate_signal now ALWAYS returns a continuous, bar-varying hammer-shape score (lower-wick fraction + close-position reward minus body and upper-wick penalties, squashed by tanh and……Show moreShow less

Iteration 2 fix for the Layer-2 'frozen signal (value=0.0)' failure across the synthetic scenarios. Root cause was the gated-signal anti-pattern: calculate_signal returned 0.0 on every bar where the strict single-bar hammer geometry did not fire, which on smooth synthetic data is every bar -> all 446 signals identical at 0.0. Fix: calculate_signal now ALWAYS returns a continuous, bar-varying hammer-shape score (lower-wick fraction + close-position reward minus body and upper-wick penalties, squashed by tanh and weighted by how far price sits above the trend SMA), so the monitoring signal varies in every scenario. The exact strict entry decision is unchanged — computed in the same place and stored in self._hammer_trigger (small body AND long lower wick AND short upper wick AND close in upper half AND close>SMA) — which should_enter and position_size now consume instead of the raw signal sign, keeping Layer-3 trade behavior and the exit logic (which already passed Layer 1) byte-identical. No imports changed (numpy only), preserving the Layer-1 pass. Venue/leverage unchanged: BINANCE futures, 4-HOUR, leverage 2.0 with position_size genuinely consuming leverage (equity * position_pct * leverage).

No edge in the base mechanism: single-bar hammer/pinbar reversal long-only on BCH 4H has negative expectancy across 2020-2026. profit_factor 0.53, avg_trade_return -439, Sharpe -3.76, and it loses in every major year (bull, bear, and chop). The edge is doubly adverse — win_rate only 41.6% AND avg_win (1204) < avg_loss (1608) — so neither hit-rate nor payoff tuning has anything to work with. The signal fires correctly (89/89 entries, 0 dropped), so this is a hypothesis-level failure, not a code or sizing bug; optimization would only overfit parameters to noise. Not worth 2 hours of optimization.

Implementation

Long-only single-bar Hammer/Pinbar reversal on BCHUSDT.BINANCE USD-M perp, 4H bars, pure OHLCV. A bar qualifies as a hammer when it has a small body (<35% of range), a long lower wick (>=2x body), a short upper wick (<=30% of body), and closes in the upper half of its range; acted on only in an uptrend (close > SMA50). calculate_signal() returns a continuous hammer-shape score (lower-wick dominance and close-position net of body/upper-wick penalties, weighted by trend position) so the signal varies every bar; the strict hammer+uptrend trigger is stored in self._hammer_trigger and consumed by should_enter/position_size. Exits: take-profit, stop-loss, trend break, or 6-bar time stop. Leverage 2.0 (~1.2x equity notional).

Verification Results

Verification failed (Layer 2 — synthetic scenarios): Parameters used: ['body_frac', 'upper_frac', 'wick_ratio', 'min_notional', 'position_pct', 'trend_period', 'max_hold_bars', 'stop_loss_pct', 'take_profit_pct'] Check that __init__ sets all attributes from self.parameters.get(). - steady_uptrend: Frozen signal: all 446 signals are identical (value=0.0). calculate_signal() likely has a bug — the signal never varies. - steady_downtrend: Frozen signal: all 446 signals are identical (value=0.0). calculate_signal() likely has a bug — the signal never varies. - flat_ranging: Frozen signal: all 446 signals are identical (value=0.0). calculate_signal() likely has a bug — the signal never varies. - volatility_spike: Frozen signal: all 446 signals are identical (value=0.0). calculate_signal() likely has a bug — the signal never varies.

Backtest Review

Sharpe
-3.76
Total return
-38.78%
Max drawdown
46.34%
Trades
89
Win rate
41.6%
Profit factor
0.53

Signal fires correctly: 89 hammer entries over 2020-2026, 0 dropped (min_notional/size), clean entry diagnostics

Reasonable trade count for the strategy type — sample is adequate, so the negative result is informative not noise

Negative expectancy: profit_factor 0.53, avg_trade_return -439, total_return -38.8%

Doubly adverse edge: win_rate 41.6% AND avg_win (1204) < avg_loss (1608)

Sharpe -3.76, Sortino -4.36, max_drawdown 46.3% — risk-adjusted performance deeply negative

Loses in every major year (2021 -25%, 2022 -9%, 2023 -14%) across all regimes — not a regime artifact

Fees+impact ~20% of gross, but strategy loses even gross — costs are secondary to the missing edge

Outcome Summary

A single-bar hammer/pinbar reversal long has no edge on BCH 4H — its losses are doubly adverse (low win rate AND average win below average loss), so the mechanism fails at the hypothesis level across all regimes and cannot be rescued by parameter tuning.

The backtest-review analyst issued an 'abandon' verdict: the signal fired correctly (89/89 entries, none dropped), so this is a hypothesis-level failure rather than a code or sizing bug — with both a sub-50% win rate and average wins smaller than average losses, neither hit-rate nor payoff tuning has anything to work with, and optimization would only overfit noise.

A long-only single-bar hammer/pinbar reversal strategy on BCHUSDT 4H Binance futures (2.0x leverage) — the portfolio's first BCH strategy and first single-bar wick-rejection mechanism — entering long after a hammer candle (small body, long lower wick, short upper wick, close in the upper half) formed as an uptrend pullback above a 50-bar SMA.

Across 89 trades over 2020-2026 it returned -38.8% with profit factor 0.53, win rate 41.6%, expectancy -$439/trade, Sharpe -3.76, Sortino -4.36 and a 46.3% max drawdown; the edge was doubly adverse (avg win $1,204 < avg loss $1,608) and it lost in every major year, with fees and impact ~20% of gross.

Backtest and paper results are hypothetical. Trading involves risk of loss.