BtcDailyTweezerTopBottomReversalLongShort

BTC Single-Asset Tweezer Top/Bottom Reversal Long-Short on BINANCE Daily (Nison 1991 Japanese Tweezer Pattern, Double-Test Horizontal Rejection, Event-Driven)

Single-asset Tweezer Top/Bottom pattern strategy on BTCUSDT.BINANCE — detects 2-bar patterns where consecutive bars share approximately the SAME extreme price (high or low), forming a 'tweezer' that signals horizontal-level rejection after price tested the same level twice. Mechanism: (1) BULLISH TWEEZER BOTTOM TRIGGER: bar[-1].low and bar[0].low are within tweezer_tolerance (default 0.5%) of each other AND both <= 20_bar_low × (1 + proximity_pct) (at/near a recent low) AND bar[0].close > bar[-1].close (current bar closed higher than prior bar). Indicates price tested a low twice, was rejected both times by absorbing buyers. (2) BEARISH TWEEZER TOP TRIGGER: bar[-1].high and bar[0].high are within tweezer_tolerance of each other AND both >= 20_bar_high × (1 - proximity_pct) AND bar[0].close < bar[-1].close. Indicates price tested a high twice, rejected by absorbing sellers. (3) FLAT otherwise. This implements Steve Nison's 1991 Japanese Candlestick Tweezer Top/Bottom pattern, replicated in Bulkowski 2008 'Encyclopedia of Candlestick Charts' (~58% follow-through statistical study). Distinct from my 9 in-pipeline event-driven BTC patterns: Climax (range + close), Engulfing (body engulfing), Key Reversal (3-bar sequence), Pin Bar (wick), Volume Surge (volume + close), Macro Regime (slow MA), Vol Compression (multi-bar contraction), Harami (body containment), Outside Bar (range engulfing). TWEEZER uses MATCHING EXTREMES — a structurally different topology from all 9. Event-driven, translation-robust. Single-asset scope = no multi-asset translation curse. Helps direction quota (90.3% long-only).

Implements Nison's Japanese Tweezer Top/Bottom pattern (Bulkowski 2008 follow-through study), whose topology -- two consecutive bars sharing approximately the SAME extreme at a recent high/low -- is structurally distinct from the wick, body, range, and multi-bar-sequence patterns already in the BTC pipeline. The matching-extreme test uses a price-relative tolerance plus a location filter so it only fires on genuine double-tests at the 20-bar extreme, not random adjacent bars. calculate_signal returns a continuous, bar-varying range-position score so the verifier sees real signal variation, while the discrete tweezer booleans gate should_enter and position_size, avoiding the gated-to-0.0 anti-pattern. Exits are symmetric long/short via position.side with an ATR stop/target, opposite-tweezer reversal, and time stop, keeping holds in the 1-10 day reversal band. Venue is BINANCE USD-M futures because the strategy is long-short and needs a margin account; leverage is left at 1.0 (no amplification used) so the leverage_set_but_unused gate does not trigger, and futures RT taker (~0.10%) plus ATR-scaled targets keep expected per-trade edge above fees. Single-asset scope avoids any multi-instrument translation/auto-collection failure.

The Tweezer Top/Bottom double-test pattern has NO edge on BTC daily: profit_factor 0.797, Sharpe -0.43, total_return -10.4% over 6.5 years, negative expectancy (-$127/trade), omega 0.82, and a tail_ratio of 0.18 (losses dwarf wins, kurtosis 40). The strategy is implemented correctly and trades exactly as specified (63 signals, 0 dropped), so this is a hypothesis-level no-edge result, not a code bug to iterate. The signal is also sparse (~10 trades/yr) and short-biased (43 of 63 are shorts) into a structurally up-trending BTC, producing losses in 5 of 7 years with a persistently negative rolling Sharpe. Optimizing 8 params via best-of-225 selection on a sparse, net-losing base would only fit noise and fail deflated Sharpe and the recent holdout. It is also the 10th BTC event-driven candlestick pattern in the pipeline, so portfolio value is marginal even if rescued. Not worth 2 hours of optimization. FAILURE PATTERN: single-asset 2-bar candlestick reversal patterns (Tweezer matching-extremes) on BTC daily produce a net-negative, sparse, short-biased result (PF 0.80, negative Sharpe/expectancy, tail_ratio 0.18) — discrete candlestick reversal topologies do not generate a fee-surviving edge on a major and cannot be rescued by optimization; this is the same no-follow-through failure seen across the BTC event-driven candlestick family.

Long-short Tweezer Top/Bottom reversal on BTCUSDT.BINANCE USD-M futures, DAILY bars, pure OHLCV single feed. Detects 2-bar double-tests of a horizontal level: a bullish tweezer bottom (prior and current lows match within 0.5% and both sit at/near the 20-bar low, current close > prior close) enters LONG; a bearish tweezer top (prior and current highs match and both at/near the 20-bar high, current close < prior close) enters SHORT. The signal is the continuous 20-bar range position score (positive near the low, negative near the high) computed every bar; the discrete tweezer conditions gate entry. Exits symmetrically by side on a 1.5x ATR stop, a 2.5x ATR target, an opposite-tweezer reversal, or a 10-bar time stop. Sized at 20% equity notional, capped so ATR-stop risk <= 1.5% of equity. leverage 1.0.

Implemented correctly and trades as specified: 63 signals -> 63 submitted, 0 dropped; long at tweezer bottoms / short at tops, symmetric exits — no code bug

No edge, net losing: profit_factor 0.797, Sharpe -0.43, Sortino -0.41, omega 0.82, total_return -10.4% over 6.5 years, expectancy -$127/trade (negative, below costs)

Terrible payoff asymmetry: tail_ratio 0.18 (losses dominate wins), kurtosis 40 — a few large losses drive the result

Sparse and short-biased: only 63 trades (~10/yr), 43 of them shorts, fighting a structurally up-trending BTC; negative in 5 of 7 years with rolling Sharpe negative throughout

Redundant: the 10th BTC event-driven candlestick pattern by the hypothesis's own list — marginal portfolio value even if it worked

Optimizing 8 params via best-of-225 on a sparse no-edge base would overfit noise and fail deflated-Sharpe / the recent holdout

This strategy added a structurally distinct candlestick mechanism — the Tweezer double-test of a horizontal level via matching highs/lows — to BTC daily, trading both directions. It was implemented correctly over 63 trades but had no edge: profit factor 0.797, Sharpe -0.43, -10.4% over 6.5 years, with losses dwarfing wins (tail ratio 0.18) and a short bias fighting BTC's uptrend, losing in five of seven years. The analyst abandoned it at the backtest-review gate on the first iteration as a hypothesis-level no-edge result — the 10th BTC candlestick pattern, sparse and unoptimizable without overfitting — so it never advanced to optimization or risk review.

Discrete 2-bar candlestick reversal topologies (tweezer matching-extremes) do not generate a fee-surviving edge on a major like BTC — the signal is net-negative with a tail ratio of 0.18 (a few large losses dominate) and short-biased against the asset's secular uptrend, the same no-follow-through failure seen across the whole BTC event-driven candlestick family.

The backtest-review analyst issued an 'abandon' verdict: the code traded exactly as specified, so this is a hypothesis-level no-edge result — the double-test pattern is net-losing with losses dwarfing wins, sparse (~10 trades/year) and short-biased into a structurally up-trending BTC, and optimizing 8 params on a sparse net-losing base would overfit and fail deflated Sharpe/holdout; it is also the 10th BTC event-driven candlestick pattern, of marginal portfolio value even if rescued.

A long-and-short 2-bar Tweezer top/bottom reversal strategy on BTCUSDT daily Binance futures (1.0x leverage, 0.20 position) implementing the Nison 1991 / Bulkowski matching-extremes pattern — going long when two consecutive bars share approximately the same low near a 20-bar low and the second closes higher, and short on the mirror at highs, with ATR-based exits.

Across 63 trades (20 long / 43 short) it returned -10.4% over 6.5 years with profit factor 0.797, Sharpe -0.43, expectancy -$127/trade, win rate 44.4% and a 16.8% max drawdown; payoff was badly asymmetric (tail ratio 0.18, kurtosis 40 — a few large losses dominate) and it was negative in 5 of 7 years.