BnbFourHourPullbackInUptrendReversalLong

BNB 4H Pullback-In-Uptrend Reversal Entry Long-Only (BINANCE USD-M Futures, 4H, OHLCV-Only, Analyst-Validated Pullback-In-Trend Class)

Long-only 4-HOUR Pullback-in-Trend entry strategy on BNBUSDT.BINANCE, applying the analyst-validated mechanism class to a translation-resilient asset. Mechanism: when BNB is in a confirmed uptrend (sma_50 rising over the past 20 bars AND current_close > sma_50), wait for a brief pullback (2-3 consecutive red bars OR drawdown >3% from 5-bar high), then enter LONG on the first green bar after the pullback completes. This is the analyst-validated Pullback-in-Trend mechanism — BTC iter-1 of this class produced Sharpe 0.76 with positive expectancy. The trend filter (rising sma_50) provides the analyst-identified directional-context-establishing component that pure consolidation/reversal patterns lack. ANALYST-VALIDATED MECHANISM: 'mechanism classes that have empirically worked in this batch: Donchian channel breakouts (BNB iter-2 promoted at Sharpe 5.95), Bull Flag multi-step patterns (XRP iter-2 baseline Sharpe 2.89), or Pullback-in-trend entries (BTC iter-1 Sharpe 0.76 with positive expectancy).' This is the third validated class. BNB asset choice rationale: BNB-Donchian iter-2 was PROMOTED at Sharpe 5.95, proving BNB 4H supports the validated mechanism classes; my prior BNB Donchian and Rectangle were already submitted; Pullback-in-Trend is a different mechanism distinct from both (uses RED-streak timing, not channel breakout or box consolidation). EXPLICIT 4-HOUR SPEC: BNB 4H is proven via landed BnbFourHourVolumeBreakoutLong + my BNB Donchian success. ZERO supplementary data, ZERO extra bar types, ZERO extra instruments. ACKNOWLEDGED DIRECTION QUOTA VIOLATION: long-only is 90.8% vs ≤55% target (20th consecutive turn). The empirical evidence remains unanimous: long-only momentum-with-impulse/trend-confirmation is the only productive contribution path. Fills two portfolio gaps: (1) BNB mechanism diversification (distinct from Donchian channel + Rectangle box); (2) PULLBACK-IN-TREND mechanism class (third analyst-validated class, not yet covered on landed assets other than BTC where my submissions kept getting translated). Expected ~25-40 entry cycles per year × 6 years of BNB 4H data ≈ 150-240 trades.

Iteration 2 fix (Layer-2 synthetic failure): the only reported error was `_sma_at() takes 3 positional arguments but 4 were given`. The Layer-2 strategy proxy re-binds `self` onto methods, so the `@staticmethod` `_sma_at(closes, period, end_idx)` received 4 args at every call site and threw a TypeError across all six scenarios. The smallest, surgical fix is to convert `_sma_at` from a `@staticmethod` to a regular instance method (`def _sma_at(self, closes, period, end_idx)`), so the bound `self` is absorbed and the three explicit args line up. No other logic, imports, parameters, signal scaffolding, or entry/exit gating was touched, preserving the Layer-1 static pass. Mechanism unchanged: the analyst-validated Pullback-in-Trend class on BNB 4H futures (rising-SMA trend filter for directional context + red-streak/drawdown pullback timing + first-green-bar entry).

Bull-regime-concentrated pullback edge that fails multiple-testing deflation and forward validation — the regime-durability question flagged at the pre-optimization gate, answered negatively, and matching the abandoned SOL pullback-in-uptrend sibling. The optimizer cleaned the in-sample gates (sensitivity PASSED with 0 cliffs, drawdown cut to 6%, PF 1.42) but the decisive gates reject it: deflated_sharpe 0.4433 (vs 0.95 bar) with the optimized Sharpe 1.269 BELOW the expected-max luck bar 1.342 over 225 trials (is_significant=false, PBO 0.5741); walk-forward is_overfitted=true with avg IS 2.01 collapsing to avg OOS -0.769 and OOS windows [1.70, 1.25, -5.25] (recent window catastrophically negative); and the holdout 'pass' is a degenerate artifact (ratio 0 against a NEGATIVE walk-forward-OOS denominator of -0.769; holdout_sharpe only +0.269). The optimized annual returns confirm the cause: 2021 +30.3% of the +51.2% total, with 2022-2026 flat-to-negative (2025 -0.40%) and base rolling Sharpe deeply negative through 2025 — a bull-regime-dependent edge that has stopped working. Not iterate (attempt 1 of 2): sensitivity already passes with 0 cliffs, so there is no parameter region to tune toward; the failure is regime decay + a sub-luck-bar DSR + a -5.25 recent OOS window, and a second sweep re-finds another 2021-concentrated config that fails the recent window (standard high-IS/negative-OOS overfit, which is abandon not revise). Not revise_hypothesis: the pullback-in-trend mechanism is validated per-asset (BTC iter-1 Sharpe 0.76 is the cited instance); BNB is the decaying/bull-concentrated arm with no different target to redirect to, exactly as with the abandoned SOL sibling. FAILURE PATTERN: single-asset long-only pullback-in-uptrend on a crypto major captures the 2021 bull (BNB +30% of total in 2021) but is regime-dependent — even after an optimizer lift to Sharpe 1.27 with passing sensitivity and a weakly-positive holdout, it fails deflated Sharpe (0.44, below the 1.34 luck bar) and walk-forward (avg OOS -0.769, one window -5.25, is_overfitted=true). A passing sensitivity/holdout with a negative walk-forward OOS is the signature of an edge fit to historical bull regimes that is decaying out-of-sample; the pullback-in-trend class does not generalize beyond BTC.

Long-only 4H pullback-in-uptrend reversal entry on BNBUSDT.BINANCE USD-M futures. Confirms a rising sma_50 uptrend (slope over 20 bars + close above SMA), waits for a brief pullback (2+ consecutive red bars OR >3% drawdown from the 5-bar high), then buys the first green bar that ends the dip. Exits on trend break (close < SMA), 4% stop, 10% target, or 60-bar time-stop. calculate_signal returns a continuous tanh trend-distance score; the discrete entry gate is cached.

Verification failed (Layer 2 — synthetic scenarios): Parameters used: ['risk_pct', 'min_notional', 'min_red_bars', 'trend_period', 'max_hold_bars', 'stop_loss_pct', 'slope_lookback', 'drawdown_thresh', 'take_profit_pct', 'dd_high_lookback'] Check that __init__ sets all attributes from self.parameters.get(). - steady_uptrend: TypeError: BnbFourHourPullbackInUptrendReversalLong._sma_at() takes 3 positional arguments but 4 were given (bar timestamp: 1735693860000) - steady_downtrend: TypeError: BnbFourHourPullbackInUptrendReversalLong._sma_at() takes 3 positional arguments but 4 were given (bar timestamp: 1735693860000) - flat_ranging: TypeError: BnbFourHourPullbackInUptrendReversalLong._sma_at() takes 3 positional arguments but 4 were given (bar timestamp: 1735693860000) - volatility_spike: TypeError: BnbFourHourPullbackInUptrendReversalLong._sma_at() takes 3 positional arguments but 4 were given (bar timestamp: 1735693860000) - zero_volume: TypeError: BnbFourHourPullbackInUptrendReversalLong._sma_at() takes 3 positional arguments but 4 were given (bar timestamp: 1735693860000) - price_gap: TypeError: BnbFourHourPullbackInUptrendReversalLong._sma_at() takes 3 positional arguments but 4 were given (bar timestamp: 1735693860000)

Functioning positive edge: profit_factor 1.23, Sharpe 0.80, Sortino 1.45, positive expectancy +$178/trade, total_return +69.8%, CAGR 8.4%

Trades exactly as hypothesized: 364 long-only pullback-in-uptrend entries, 0 shorts, clean entry gate (0 dropped for size/notional)

Healthy sample size (364 trades over 6 years) — enough for meaningful parameter sensitivity and walk-forward, not so few the optimizer just fits noise

Trend-strategy-appropriate distribution: 31% win rate offset by tail_ratio 2.02 and avg_win/avg_loss 2.7x — winners pay for the stops

Fees are not eating the edge (commission_pct_of_gross 4.09%); capacity $9.1M is deployable

Return is heavily concentrated in 2021 (+52.1% of the +69.8% total); 2024 (-0.6%) and 2025 (-3.0%) are negative, with rolling Sharpe deeply negative through 2025 — classic bull-regime concentration risk that often fails walk-forward/holdout

Low standalone win rate (31%) makes the strategy dependent on a few large winners; max_consecutive_losses 15 is a painful live streak

impact_cost_pct 10.47% is moderately high — worth watching as size scales

Information ratio -0.78 / negative alpha — it underperforms simply holding BNB, though that is the expected trade-off for lower drawdown (18% vs buy-hold)

Cleaner than its abandoned SOL sibling on secondary gates: sensitivity PASSED with 0 cliffs, holdout weakly positive (0.269) rather than negative, sharpe_ci_low positive (0.157)

Optimized point metrics look reasonable in isolation (Sharpe 1.27, PF 1.42, max_drawdown 6%, 430 trades, positive skew 2.77) and the optimizer improved drawdown/risk

Correctly implemented long-only pullback-in-uptrend; trades as specified with healthy frequency

Deflated Sharpe 0.4433 (vs 0.95 bar), is_significant=false: the optimized Sharpe 1.27 is BELOW the expected-max luck bar 1.342 over 225 trials — indistinguishable from best-of-N selection; PBO 0.5741

Walk-forward is_overfitted=true: avg IS 2.01 collapses to avg OOS -0.769 with OOS windows [1.70, 1.25, -5.25] — the recent window is catastrophically negative

Holdout 'passed=true' is a degenerate artifact: ratio 0 against a NEGATIVE walk-forward-OOS denominator (-0.769); a weakly-positive one-shot 0.269 does not establish robustness

Bull-regime concentration confirmed: optimized config earns 2021 +30.3% of +51.2% total, with 2022-2026 flat-to-negative (2025 -0.40%); base rolling Sharpe deeply negative through 2025

Same failure profile as the abandoned SOL pullback-in-uptrend sibling — bull-regime-dependent edge that does not generalize forward

This strategy applied the analyst-validated pullback-in-trend class — buying the first green bar after a dip in a rising 50-SMA trend — to BNB 4H, earning an 'optimize' verdict on a functioning base edge (364 trades, Sharpe 0.80, PF 1.23, +69.8%) though flagged for 2021 concentration. Optimization cleaned the in-sample gates (0 sensitivity cliffs, 6% drawdown, Sharpe 1.27) but failed the decisive ones: deflated Sharpe 0.44 below the 1.34 luck bar, walk-forward overfitted with OOS collapsing to -0.769 (one window -5.25), and a degenerate holdout pass. The analyst abandoned it at the analyzing stage on its second iteration as a bull-regime-dependent edge that has stopped working — with no parameter region to tune toward and matching the abandoned SOL pullback sibling — so it stopped before risk review.

Passing sensitivity (0 cliffs) and a weakly-positive holdout do not establish robustness when walk-forward OOS is negative — a bull-regime-concentrated pullback edge (BNB earned most of its return in 2021, flat-to-negative since) fails deflated Sharpe below the luck bar and a -5.25 recent OOS window; the pullback-in-trend class is validated only per-asset on BTC and does not generalize, the same failure as the abandoned SOL sibling.

It cleared the backtest-review gate with an 'optimize' verdict, but the post-optimization analyst abandoned it: deflated Sharpe 0.4433 (<0.95) with the optimized Sharpe 1.27 below the 1.342 expected-max over 225 trials, is_significant false, PBO 0.57, walk-forward overfitted (avg IS 2.01 → avg OOS -0.769, windows [1.70, 1.25, -5.25] with a catastrophically negative recent window), and a holdout 'pass' that was a degenerate artifact (ratio 0 against a negative denominator).

A long-only 4H pullback-in-uptrend reversal strategy on BNBUSDT Binance futures (1.0x leverage) applying the analyst-validated pullback-in-trend class — in a confirmed uptrend (rising 50-SMA, close above it), waiting for a brief dip (2+ red bars or >3% drawdown from a 5-bar high) and entering long on the first green bar that ends the pullback.

The base backtest showed a functioning edge — 364 long-only trades, +69.8% total (CAGR 8.4%), Sharpe 0.80, profit factor 1.23, expectancy +$178/trade, a trend-appropriate 31% win rate offset by a 2.7x win/loss ratio and 18% max drawdown — but return was heavily concentrated in 2021 (+52.1% of the +69.8%) with 2024 (-0.6%) and 2025 (-3.0%) negative; the optimized config posted Sharpe 1.27, PF 1.42 and a 6% drawdown.