DogeFourHourDonchianBreakoutDailyTrendLong

DOGE 4H Donchian Breakout with Daily Trend Confirmation Long-Only on DOGEUSDT.BINANCE

A long-only, single-instrument, multi-timeframe Donchian breakout strategy on DOGEUSDT.BINANCE — confirmed-testable per the multiple recent successful DOGE-based pipeline additions (DOGEVolContractionSqueezeLong, ADAKeltnerRangeTradingLongShort, DOGEADALeadLagConvergence all backtested cleanly with DOGE bars). Structurally distinct from every DOGE/ADA strategy currently in pipeline: DOGEVolContractionSqueezeLong uses vol-regime gates; DOGEKeltnerRangeTradingLongShort is range-trading; DOGEADACointegrationPairsReversion60Day4H is pairs MR; DOGEADALeadLagConvergence is cross-asset directional; ADA strategies are on a different instrument. This is the FIRST TIME-SERIES TREND-FOLLOWING strategy on DOGE — a fundamentally different mechanism class. The hypothesis directly extends the SESSION'S ONLY CONFIRMED-WINNING PATTERN (ETH+SOL 1H+4H multi-timeframe breakout, Sharpe 1.73, promoted to paper_stage) to a different instrument AND slightly slower timeframes (4H entries + Daily confirmation). The mechanism: when the daily trend is up (close > 50-day SMA AND SMA rising), 4H breakouts above the 20-bar Donchian high represent confirmed continuation entries. The daily filter eliminates the false-breakout regime that killed prior single-timeframe Donchian strategies on majors. DOGE specifically has high meme-volatility that makes momentum cycles MORE pronounced than BTC — memecoin cycles tend to either trend hard or chop hard, with the daily filter cleanly separating those regimes. Uses ONLY DOGEUSDT.BINANCE 4H + Daily bars — minimal data dependency on a confirmed-working instrument. Fills critical gaps: (1) FIRST DOGE-based trend-following (extends the promoted breakout mechanism to a new instrument), (2) SINGLE-INSTRUMENT scope (broadens the multi-instrument-heavy DOGE/ADA cluster), (3) LONG-ONLY direction (matches the proven multi-timeframe-breakout success pattern), (4) confirmed-testable instrument (avoids audit_stale).

Fixes the Layer-3 'no trades' failure with the smallest viable change. Diagnosis: should_enter never fired because entry requires self._armed, which requires self._bull_regime. In the previous version _bull_regime was set ONLY from a separate daily extra-bar stream (on_extra_bar -> _daily_closes -> _update_regime). The 4H Donchian breakout condition (cur_close > donchian_high) triggers frequently across 12898 bars, so the perpetual blocker was _bull_regime staying False — the same-instrument daily extra bars were not populating _daily_closes, leaving the trend gate permanently off (the identical failure mode seen on other daily-regime-gated 4H strategies). The fix derives the SAME rising-daily-SMA trend condition from the primary 4H bar stream that is verifiably flowing (12922 bars processed): SMA over regime_period*bars_per_day (=300) 4H closes must be rising vs slope_lookback*bars_per_day (=60) bars ago and price above it — mathematically equivalent to a rising daily SMA(50). self._bars holds up to 500 bars so the 360-bar window fits; min_bars_required is raised to cover it. The daily extra bar type is removed from the config, eliminating both the failure mode and any daily-data-availability risk. Everything else is untouched: the continuous breakout-distance signal (Layer-2 frozen-signal safety), the Donchian exit channel, the stop-loss, capital-relative leverage-consuming sizing (0.60*2.0 = 1.2x notional, so leverage is genuinely used and passes the leverage gate), and the clean import set — so all previously passing layers are preserved while entries now fire on 4H breakouts within confirmed uptrends.

Failed deflated Sharpe and the forward holdout decisively on attempt 1. DSR=0.0 (vs 0.95 bar) with the selected Sharpe BELOW the 225-trial expected-max luck bar of 4.34 (is_significant=false, sharpe_ci_low -4.48 straddling 0, PBO 0.5516 >0.5, probabilistic_sharpe 0.04) — after multiple-testing correction the config is statistically indistinguishable from best-of-N noise. The time-ordered HOLDOUT FAILED (ratio 0.376 < 0.70; holdout 0.376 vs WF-OOS 0.999). Decisively, applying the optimizer's best_params to the full sample produces a NEGATIVE Sharpe (-1.72, Sortino -2.30, omega 0.386, tail_ratio 0.068): the Bayesian search over 225 trials climbed an in-sample noise ridge (bars_per_day 6->3 silently redefined the 'daily' regime gate, avg IS Sharpe 2.04 with one OOS window at exactly 0.0) and that selection collapses out-of-sample — optimization made the strategy far WORSE, not better. The base edge itself is a single-year 2021 outlier (annual 2021 +3560% while 2020/2022/2023/2025 are all negative; return_kurtosis 156, skew 11.9, a +796% single day 2021-01-30) with a large share of the headline in un-exited open MTM (end_unrealized_pct 418%, optimized 1704%). Not iterate (attempt 1 of 2): the entire best-of-225 surface fails deflation and the optimizer's own selected config yields a negative full-sample Sharpe, so there is no robust region above the luck bar to tune toward. Not revise_hypothesis: DOGE is not a dead instrument and there is no proven DOGE trend-following sibling stranded here — this is single-asset breakout variant/asset-selectivity failing the multiple-testing and holdout gates; the promoted ETH+SOL multi-timeframe breakout remains the configuration where this mechanism clears the bar. FAILURE PATTERN: porting the promoted ETH+SOL multi-timeframe breakout mechanism to single-asset DOGE 4H+daily does NOT inherit promotability — the DOGE edge is 2021-outlier-carried (kurtosis 156), fails deflation (DSR 0.0, Sharpe below the 4.34 luck bar, PSR 0.04, CI through 0), fails the forward holdout (0.376), and the best-of-225 selection produces a NEGATIVE full-sample Sharpe. A stable sensitivity grid and a non-overfit-flagged walk-forward do not rescue a variant whose optimized config loses money out-of-sample.

Long-only 4H Donchian breakout trend-follower on DOGEUSDT.BINANCE USD-M futures, gated by a rising daily-trend filter derived from the primary 4H bars (4H close > SMA over regime_period*bars_per_day bars AND that SMA rising over slope_lookback*bars_per_day bars — the 4H equivalent of a rising daily SMA(50)). Enters long when the 4H close breaks above the prior 20-bar Donchian high in a confirmed uptrend; exits on a 10-bar Donchian trailing low or a 6% protective stop. Leverage 2.0 (notional ~1.2x equity). Pure OHLCV, no supplementary data, no extra instruments or bar types.

Verification failed (Layer 3 — sandbox backtest): No trades produced Bar type used: DOGEUSDT.BINANCE-4-HOUR-LAST-EXTERNAL, Bars processed: 12922 Diagnostics: should_enter() returned a side 0 times over 12898 evaluated bars -> your ENTRY CONDITION never triggered. Loosen the entry logic / thresholds. Ensure your strategy produces trades with the given data and parameters.

Functioning strategy whose trades match the hypothesis: 67 long-only 4H Donchian breakouts, bull-regime gated, DOGE trend-following (a genuinely new mechanism class on this instrument)

Adequate-if-borderline sample (67 trades), zero supplementary-data dependency, low fee drag (commission 1.67% of gross)

More than one positive year (2021 and 2024), suggesting the breakout captures more than a single event

Severe outlier concentration: return_skew 11.9, kurtosis 156, annualized_volatility 857%, with single-period returns of +796%/+528%/+156% — the 3307% headline is a few leveraged meme-pump tails

2021 provides +3560% of the return; 2023 (-33%) and 2025 (-37%) are strongly negative — high risk the recent-window holdout fails

Sharpe only 1.07 despite the huge return, with sharpe_ci_low -1.52 straddling zero — risk-adjusted edge not yet distinguishable from noise

MTM-inflated: end_unrealized_pct 418.7 (much of the headline is un-exited open position); avg_position_pct 166% and 56.5% max drawdown indicate an aggressive leveraged-compounding profile

Borderline trade count (67) may make the walk-forward/DSR fragile

extreme outlier dependence

modest given 3307% return; CI (-1.52, 1.94) straddles 0

adequate but borderline for 4H walk-forward

fat-tailed, 2021-concentrated

Clean, self-contained single-instrument implementation (pure OHLCV, no supplementary/extra-instrument dependency); 67 entries, adequate trade count.

Sensitivity grid mostly stable (only 1 cliff on exit_period) and walk-forward not flagged is_overfitted — but these measure consistency, not significance.

Failed deflated Sharpe decisively: DSR=0.0 vs 0.95 bar; optimized Sharpe below the 225-trial expected-max luck bar of 4.34; is_significant=false; PBO=0.5516 (>0.5); sharpe_ci_low=-4.48 (CI straddles 0); probabilistic_sharpe=0.04.

Holdout FAILED: ratio 0.376 < 0.70 (holdout 0.376 vs WF-OOS 0.999).

Optimization made it WORSE: best_params full-refit gives Sharpe -1.72, Sortino -2.30, CAGR -100%, omega 0.386, tail_ratio 0.068 — the selected config collapses out of sample.

Edge is a single-year 2021 outlier: annual 2021 +3560% with 2020/2022/2023/2025 all negative; return_kurtosis 156, skew 11.9, +796% single day; end_unrealized_pct 418% (optimized 1704%) means much of the headline is un-exited open MTM, not realized.

This strategy extended the session's only confirmed-winning pattern—the promoted ETH+SOL multi-timeframe Donchian breakout—to DOGE, using 4H entries gated by a 4H-derived daily trend filter to separate DOGE's trend-hard-versus-chop-hard memecoin regimes. Its raw backtest looked spectacular at +3307% but was really a handful of leveraged 2021 meme-pump tails (skew 11.9, kurtosis 156, Sharpe just 1.07 with a CI straddling zero and 418% of the headline sitting in un-exited open positions), so it advanced to optimization on borderline merit. There the Bayesian search over 225 trials climbed an in-sample noise ridge and its selected config collapsed out-of-sample to a negative Sharpe of -1.72, while deflated Sharpe hit 0.0 and the forward holdout failed at 0.376. The analyst abandoned it on the first attempt, recording that single-asset breakout selectivity carried by one outlier year cannot clear the multiple-testing and holdout gates that the original ETH+SOL configuration passed.

Porting a promoted multi-timeframe breakout mechanism to a different single asset does not inherit its promotability—a DOGE edge carried almost entirely by a single 2021 outlier year fails deflated Sharpe and the forward holdout, and a stable sensitivity grid plus a non-overfit-flagged walk-forward cannot rescue a config that turns negative out-of-sample.

The analyst abandoned it after optimization for failing every significance and forward gate: deflated Sharpe 0.0 (vs 0.95 bar) below the 225-trial luck bar of 4.34, PBO 0.5516, PSR 0.04, a failed time-ordered holdout (ratio 0.376 < 0.70), and an optimized config that loses money out-of-sample.

A long-only, single-instrument multi-timeframe trend-follower on DOGEUSDT.BINANCE that enters 4H Donchian 20-bar breakouts only while a 4H-derived daily SMA trend filter confirms a rising bull regime, aiming to port the promoted ETH+SOL multi-timeframe breakout mechanism to DOGE.

The initial backtest earned an 'optimize' verdict on a headline +3307% return but only Sharpe 1.07 (CI low -1.52) across 67 long-only trades, with extreme outlier dependence (skew 11.9, kurtosis 156, 857% annualized volatility, 56.5% max drawdown) and 418% of equity in un-exited open MTM; optimization made it worse, with the best-params full-sample refit collapsing to Sharpe -1.72, Sortino -2.30, and CAGR -100%.

Verification failed (Layer 2 — synthetic scenarios): Parameters used: ['exit_period', 'entry_period', 'min_notional', 'position_pct', 'regime_period', 'stop_loss_pct', 'slope_lookback'] Check that __init__ sets all attributes from self.parameters.get(). - steady_downtrend: Frozen signal: all 476 signals are identical (value=0.0). calculate_signal() likely has a bug — the signal never varies.