BnbDailyDonchianBreakoutLong

BNB Daily Volume-Confirmed Donchian Breakout Long (Pure Mechanism)

A long-only, single-instrument, single-timeframe strategy on BNBUSDT.BINANCE DAILY bars implementing the PURE analyst-validated Volume-Confirmed Donchian Breakout template. Fills a clean instrument-timeframe gap: BnbFourHourVolumeBreakoutLong exists on 4H but no BNB Daily strategy exists, and the existing 4H entry may use a different volume-breakout mechanism (volume breakout vs explicit Donchian breach). The PURE template: 20-day Donchian high breakout + volume confirmation + 50-day SMA trend gate. Three filters only — keeps trade count high for walk-forward statistical power per the analyst's recurring guidance ('over-filtered strategies fail walk-forward'). The recent abandons of BTC 4H Anti-Chop and XRP 4H Rising-Trend confirm the analyst's diagnosis: adding regime filters compounds the rare-event walk-forward methodology mismatch. The pure 3-filter template has consistently produced strong default backtests (Sharpe 2-3, 6/7 positive years on ADA 4H; expected similar profile on BNB Daily). BNB Daily provides three specific advantages: (a) BNB is the Binance exchange token with the most institutional stability of any non-BTC/ETH crypto — deep liquidity, less idiosyncratic noise; (b) BNB has distinct narrative cycles from BTC/ETH/SOL/ADA/LINK/XRP — exchange-volume cycles, BSC ecosystem activity, BNB burn schedules; (c) BNB Daily bars are CONFIRMED WORKING via BnbFourHourVolumeBreakoutLong (which validates the BNB instrument; Daily timeframe on the same Binance Vision data source).

Implements the analyst-validated pure 3-filter Volume-Confirmed Donchian Breakout template on the open instrument-timeframe gap (BNB Daily; only BNB 4H exists). calculate_signal returns the continuous breakout distance — (close - prior 20-day high)/prior high — which varies every bar and is positive exactly when price breaks the channel, so the signal carries conviction rather than being gated to a constant; should_enter thresholds it (>0) and layers the volume-confirmation and 50-day-SMA trend gates. The Donchian high and volume average both exclude the current bar so the breakout candle is judged against a clean prior reference. The turtle lower-channel exit keeps turnover and trade count high (walk-forward statistical power, per the analyst's 'over-filtered strategies fail walk-forward' guidance) while letting trends run. VENUE: the strategy is long-only, no leverage, OHLCV momentum (not funding), so per the venue-selection rule it routes to BINANCE_SPOT (CASH); leverage stays 1.0 and is intentionally unused in sizing (spot caps at 1x), avoiding the leverage_set_but_unused rejection. BNB is deeply liquid with long daily Binance history. Daily frequency (~20-40 trades/yr) makes the ~0.20% spot round-trip fee negligible against multi-percent breakout moves.

Failed deflated Sharpe: DSR=0.2213, expected-max=7.26 over 225 trials — the optimized Sharpe of 5.82 is below the best-of-N luck bar and not significant (is_significant=false). Although two robustness signals are favorable in isolation (sharpe_ci_low 1.53 > 0 and PBO 0.4577 <= 0.5), they are full-period bootstrap/CSCV statistics that do NOT correct for the 225-trial selection the way DSR does, and they are overwhelmed by the decisive failures: walk-forward is_overfitted=true with avg IS 7.71 collapsing to avg OOS 0.0 and ALL three OOS windows exactly [0.0, 0.0, 0.0] — the optimizer pushed vol_mult to 2.01 / sma_period to 25, over-selecting so the chosen config produces ZERO out-of-sample trades in every window — plus a degenerate holdout (Sharpe 0.0 against a 0.0 WF-OOS reference, ratio 0). A config that produces no out-of-sample trades has not generalized regardless of its in-sample CI. The optimized headline (2228%, annual 2018 +200% from a +221% single-day attribution, 2021 +172%) confirms the single-trade/bull concentration flagged at pre-optimization (95.7% of P&L in one 2021 trade). This is overfit by over-selectivity, not a tunable-region case: the 225-trial sweep's best config does not trade out-of-sample, so a second pass over the same space hits the identical wall — there is no robust region to tune toward (sensitivity passing on the trivial risk_fraction>1.0 cliff is necessary but not sufficient). Not revise_hypothesis: the volume-breakout family is validated on its other instruments (BNB 4H, ADA 4H), so BNB Daily is a dead target rather than a premise needing reframing. FAILURE PATTERN: 7-parameter volume-confirmed Donchian breakout on a single alt whose edge is one bull-cycle trade — the optimizer maximizes in-sample Sharpe by raising vol_mult to over-select, yielding zero walk-forward OOS trades and a degenerate holdout; favorable full-period sharpe_ci_low/PBO do not override DSR<<0.95 and all-zero OOS.

Long-only single-instrument volume-confirmed Donchian breakout on BNBUSDT (spot), DAILY bars, OHLCV-only. Enters long when the close breaks above the prior 20-day high, today's volume is at least 1.5x the prior 20-day average volume, and the close is above the 50-day SMA. Exits via a turtle channel exit (close below the prior 10-day low), an 8% protective stop, or the close falling back below the 50-day SMA — no take-profit or time stop so winners run. Sized to 90% of equity, long-only, no leverage.

Strategy functions correctly and matches the hypothesis: 3-filter Donchian breakout (20-day high + volume confirm + 50-day SMA gate) with turtle-channel/stop exits. Dominant winner is a real 96-day BNB 2021 breakout, not an artifact.

Stronger profit factor (1.98) than the thin-PF (~1.1) volume-breakout ports that have repeatedly failed; 12 of 32 trades win.

Recent regime is POSITIVE (2025 +26%), unlike the regime-hold siblings whose holdout windows were dead; 5 of 8 full years positive (2019/2020/2021/2024/2025).

Low end_unrealized_pct (16%) — not an open-position mark-to-market mirage; 32 trades adequate for a sweep.

Extreme single-trade concentration: the 2021 +602% trade is 95.7% of total P&L. Strip it and the strategy is roughly break-even — the same one-trade dependence that has failed DSR/holdout across this session's siblings.

7 tunable parameters — the largest overfit/cliff surface of the batch; high risk of sensitivity cliffs and elevated PBO in optimization.

max_drawdown 67% with a 1680-day (4.6yr) max drawdown duration; negative 2018/2022/2023/2026 years.

Headline Sharpe (3.34) distorted by per-day whole-trade attribution (annualized_volatility 1588%, sharpe_ci_low -4.29 straddles 0); unreliable until the robustness block is computed.

overfit-surface risk

adequate

1 trade = 95.7% of P&L

watch

yes

must match

Clean implementation matching the hypothesis (3-filter volume-confirmed Donchian breakout); sensitivity effectively clean (only the trivial risk_fraction>1.0 cliff).

Some robustness signals are favorable in isolation: sharpe_ci_low 1.53 (positive) and PBO 0.4577 (<=0.5) — but they are full-period statistics that don't survive the multiple-testing and OOS gates.

Deflated Sharpe 0.2213 vs the 7.26 best-of-225 luck bar — optimized Sharpe 5.82 below it; is_significant=false.

Walk-forward is_overfitted=true with avg IS 7.71 collapsing to avg OOS 0.0 and ALL three OOS windows [0.0, 0.0, 0.0] — the selected config (vol_mult 2.01) produces zero out-of-sample trades.

Holdout degenerate-fail (Sharpe 0.0 vs 0.0 WF-OOS reference, ratio 0).

Edge remains single-trade/bull-concentrated: optimized annual 2018 +200% (a +221% single-day attribution), 2021 +172%; pre-opt flagged 95.7% of P&L in one 2021 trade.

expected-max 7.26

BnbDailyDonchianBreakoutLong applied the analyst-endorsed pure 3-filter volume-confirmed Donchian breakout template to the BNB daily gap, with a healthier profit factor (1.98) and a positive recent regime than its session siblings. It cleared the pre-optimization gate but reviewers flagged that one 2021 trade was 95.7% of P&L and that 7 parameters posed a large overfit surface. Optimization confirmed the concern: deflated Sharpe 0.22, an in-sample 7.71 collapsing to zero OOS activity as the optimizer over-selected on vol_mult, and a degenerate holdout — overfit by over-selectivity with no robust region to tune toward, and the favorable CI/PBO unable to rescue it. The analyst judged BNB Daily a dead target for the validated family and abandoned it; it reached optimization and analysis but was abandoned after a single iteration, never advancing to risk review.

Favorable full-period statistics like a positive sharpe_ci_low and sub-0.5 PBO do not override a deflated Sharpe far below the best-of-N luck bar and all-zero walk-forward OOS: a 7-parameter volume-Donchian breakout whose edge is one bull-cycle trade gets 'optimized' by raising vol_mult until the config produces zero out-of-sample trades — the family is validated on other instruments (BNB 4H, ADA 4H), so BNB Daily is a dead target.

After passing the backtest-review gate (verdict: optimize), the analyst abandoned it at the post-optimization stage: it failed the deflated Sharpe test (DSR 0.22 vs expected-max luck bar 7.26 over 225 trials, is_significant=false) and the walk-forward was overfit (avg IS 7.71 collapsing to all-zero OOS windows because the over-selective optimized config never traded out-of-sample), with a degenerate holdout (ratio 0) — even though sharpe_ci_low (1.53) and PBO (0.46) were favorable in isolation.

A long-only volume-confirmed Donchian breakout strategy on BNBUSDT.BINANCE daily bars (OHLCV only) using the pure 3-filter template — a 20-day Donchian high breakout, volume confirmation (≥1.5× the prior 20-day average), and a 50-day SMA trend gate — exiting on a 10-day turtle channel low, an 8% stop, or a close back below the SMA, filling the BNB-daily gap left by the existing 4H volume-breakout strategy.

The base backtest over 2017-11-06 to 2026-05-18 showed strong-looking but concentrated results across 32 trades — total return +890%, CAGR 31.4%, profit factor 1.98, Sharpe 3.34 (CI [-4.29, 6.20]), a 67% max drawdown, 37% win rate — with one 2021 trade accounting for 95.7% of total P&L; optimization (Sharpe 5.82, +2228%) raised vol_mult and produced zero out-of-sample trades.