BnbDailyMultiWeekTrendContinuationLong

BNB Daily Multi-Week Trend Continuation Long-Only (BINANCE USD-M Futures, Daily, OHLCV-Only)

Long-only BNB daily strategy on BINANCE USD-M futures. Same proven multi-week trend-continuation template as the landed Link/Btc/Sol daily versions — applied to BNB to extend the mechanism family to a fourth instrument while keeping single-instrument + OHLCV-only + factory-approved (BNBUSDT.BINANCE 1D) discipline. The already-landed BnbFourHourVolumeBreakoutLong proves BNB has clean continuation behavior; this is the daily-timeframe analog.

Iteration 2 fix: Layer 2 (synthetic) failed with `_ema() takes 2 positional arguments but 3 were given` because `_sma` and `_ema` were declared as `@staticmethod`. The Layer-2 synthetic proxy invokes strategy helpers as self-receiving bound methods, so the 2-arg staticmethods received 3 args. The minimal fix converts `_sma` and `_ema` from staticmethods to instance methods (add `self`, drop `@staticmethod`) — matching `_compute_atr` which already worked under the same proxy. All call sites already use `self._sma(...)`/`self._ema(...)`, so no other change was needed. Signal logic, parameters, sizing, and imports are byte-identical to the previous version, preserving every Layer 1 check that already passed.

Do not optimize — this exact strategy (BnbDailyMultiWeekTrendContinuationLong) was already abandoned post-optimization with DSR 0.0 (vs a 4.19 expected-max luck bar), walk-forward is_overfitted with ALL three OOS windows producing ZERO trades, a zero-trade holdout, and PBO 0.73. The re-submitted base backtest carries the identical disqualifying signature: the entire +126% return rests on ~2 outlier days (2021-03-06 +88.3%, largest_win $86k; 2024-04-02 +17.3%) with return_kurtosis 571 and skew 23.5 — strip them and every year is flat-to-negative (2020 -2.5%, 2022 +1.1%, 2023 -3.9%, 2025 +4%), and sharpe_ci_low is -1.75 (CI straddles 0), so Sharpe 0.64 / PF 3.99 / Sortino 8.2 are pure single-day artifacts. The cause is structural: the 4-condition AND-gate (200-SMA regime + 30/80 EMA stack + 20-day momentum + 5-day-high trigger) fires only 30 times in 6 years, which starves the walk-forward OOS windows (the documented zero-trade-OOS failure). This is the same single-outlier sparsity mode as the abandoned SOL (19 trades, 2-day edge) and ADA (15 trades, 1-day edge) daily multi-week-continuation siblings. The trend-continuation mechanism is sound but already landed on the targets where it works (LINK/BTC/SOL daily + BnbFourHourVolumeBreakout on BNB); BNB *daily* multi-week continuation is the sparse, outlier-dependent arm that cannot be validated out-of-sample, and no parameter optimization can manufacture a distributed edge from a 2-day artifact.

Long-only BNBUSDT daily multi-week trend-continuation strategy on BINANCE USD-M futures. Enters when (1) regime: close > 200-SMA, (2) trend stack: EMA30 > EMA80, (3) momentum: 20-day return > 0, AND (4) continuation trigger: close breaks the prior 5-day high. Exits on ATR(14) x3 chandelier trail break, trend break (close < EMA80), or regime break (close < 200-SMA). Capital-relative ATR-risk sizing at 2% risk per trade.

Verification failed (Layer 2 — synthetic scenarios): Parameters used: ['atr_mult', 'ema_fast', 'ema_slow', 'risk_pct', 'atr_period', 'mom_period', 'sma_period', 'min_notional', 'size_precision', 'breakout_period', 'max_notional_frac'] Check that __init__ sets all attributes from self.parameters.get(). - steady_uptrend: TypeError: BnbDailyMultiWeekTrendContinuationLong._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735701840000) - steady_downtrend: TypeError: BnbDailyMultiWeekTrendContinuationLong._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735701840000) - flat_ranging: TypeError: BnbDailyMultiWeekTrendContinuationLong._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735701840000) - volatility_spike: TypeError: BnbDailyMultiWeekTrendContinuationLong._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735701840000) - zero_volume: TypeError: BnbDailyMultiWeekTrendContinuationLong._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735701840000) - price_gap: TypeError: BnbDailyMultiWeekTrendContinuationLong._ema() takes 2 positional arguments but 3 were given (bar timestamp: 1735701840000)

Mechanism is sound and already validated/landed on the appropriate targets (LINK/BTC/SOL daily multi-week continuation; BnbFourHourVolumeBreakout on BNB itself)

Clean low-cost mechanics (commission 0.68% of gross, max_DD 14.6%), no correctness bugs

Entire return is ~2 outlier days: 2021-03-06 +88.3% (largest_win $86k) and 2024-04-02 +17.3%; return_kurtosis 571, skew 23.5 — ex-outlier the record is flat-to-negative every year

sharpe_ci_low -1.75 (CI deeply straddles 0); Sharpe 0.64 is indistinguishable from noise; PF 3.99 / Sortino 8.2 are outlier-inflated

Structural sparsity: 30 trades over 6 years from a 4-condition AND-gate; split across 3 walk-forward windows this starves the OOS windows

This EXACT strategy was already abandoned post-optimization (DSR 0.0 vs a 4.19 luck bar, is_overfitted with all 3 OOS windows zero-trade, zero-trade holdout, PBO 0.73) — re-running optimization will reproduce that result

Same single-outlier sparsity failure mode as the abandoned SOL and ADA daily multi-week-continuation siblings

BnbDailyMultiWeekTrendContinuationLong attempted to extend the validated daily multi-week trend-continuation family to a fourth instrument, BNB, with a clean single-instrument futures design. Its base backtest looked superficially strong (126.5% return, PF 3.99), but the reviewer identified that the entire edge came from roughly two outlier days, leaving a Sharpe of 0.64 with a CI straddling zero and a return distribution with kurtosis 571. Because the four-filter AND-gate fired only 30 times in six years — the same single-outlier sparsity mode that sank the SOL and ADA daily siblings — and because this exact strategy had already failed optimization once (DSR 0.0, zero-trade OOS windows, PBO 0.73), the analyst abandoned it at the backtest-review gate without re-running optimization.

A 4-condition AND-gate that fires only 30 times in 6 years cannot manufacture a distributed, out-of-sample-validatable edge — when the headline return rests on one or two outlier days, no parameter optimization can rescue it, and re-submitting an already-abandoned config reproduces the same failure.

The analyst issued an 'abandon' verdict at the pre-optimization backtest-review gate, so optimization and the post-optimization analysis were never reached; the decisive reason was that this exact strategy had already been abandoned post-optimization (DSR 0.0 vs a 4.19 luck bar, walk-forward overfit with all three OOS windows zero-trade, zero-trade holdout, PBO 0.73), and the resubmitted base carried the identical single-outlier, structurally sparse signature.

A long-only, single-instrument, OHLCV-only BNBUSDT daily futures strategy porting the LINK/BTC/SOL multi-week trend-continuation template to BNB — entering on a 5-day-high breakout only when all of a 200-SMA regime filter, a 30/80 EMA trend stack, and positive 20-day momentum hold, then trailing out via an ATR chandelier, EMA-slow break, or regime break.

The base backtest showed 126.5% total return (CAGR 15.9%) over just 30 trades with profit factor 3.99 and a 14.6% drawdown, but Sharpe was only 0.64 with a CI low of -1.75 and extreme tails (kurtosis 571, skew 23.5) — the reviewer found the entire return rested on roughly two outlier days (2021-03-06 +88.3% and 2024-04-02 +17.3%), with every year flat-to-negative once those are stripped out.