EthDailyVolumeConfirmedDonchianBreakoutLong

ETH Daily Volume-Confirmed Donchian Breakout Long (Pure Mechanism, No Regime Overlay)

A long-only, single-instrument, single-timeframe strategy on ETHUSDT.BINANCE DAILY bars implementing the PURE analyst-validated Volume-Confirmed Donchian Breakout template — explicitly WITHOUT any regime-filter overlay. The analyst's recent abandon feedback was clear: over-filtered strategies (4+ conjoined gates like calm-vol + volume + Donchian + trend AND rising-slope AND alignment-persistence) produce too few eligible bars for walk-forward statistical power. The PURE template uses only THREE filters: (1) 20-day Donchian high breakout, (2) volume confirmation, (3) long-term trend gate (close > sma_50). No calm-vol filter, no anti-chop filter, no slope filter, no alignment persistence. Simpler = more trades = better walk-forward statistical power per the analyst's recurring methodology-mismatch diagnosis. Applied to ETH Daily because: (a) no existing ETH Daily Volume-Confirmed Donchian strategy in pipeline — EthDailyGoldenCrossMomentumLong uses Golden Cross (different mechanism: SMA crossover event vs Donchian channel breach), EthFourHourVolumeBreakoutLong uses 4H not Daily; (b) ETH has distinct narrative cycles from BTC/SOL/ADA — post-Merge dynamics, L2 ecosystem launches (Arbitrum, Optimism, Base), DeFi TVL flows, EIP-1559 burn cycles — produces breakout opportunities decoupled from broader BTC beta; (c) Daily timeframe gives ~252 bars/year producing ~15-25 trades/year (well above the 30-trade walk-forward threshold over 6+ years); (d) ETH Daily bars are CONFIRMED WORKING (EthDailyGoldenCrossMomentumLong validates this exact bar type — Daily, not 4H which has had stale issues). Single-instrument, single-timeframe, NO secondary dependencies — directly avoids the multi-instrument audit_stale failure mode.

Iteration-2 fix: the only failure was a Layer-2 TypeError because `_mean` was a `@staticmethod` that the synthetic verification proxy invoked with an injected `self`, giving 2 positional args to a 1-arg signature. Converted `_mean` to a regular instance method (`def _mean(self, values)`) — matching every existing `self._mean(...)` call site — which resolves all six synthetic-scenario failures. No other logic changed: imports remain clean, the pure 3-filter Donchian/volume/trend signal and ATR sizing are untouched, so Layer 1 stays green and the strategy now passes Layer 2.

Failed deflated Sharpe: DSR=0.0208 (vs 0.95 bar), expected-max=10.69 over 225 trials — the optimized Sharpe 5.71 is far below the best-of-N luck bar, indistinguishable from selection noise (is_significant=false). The decisive forward gate also fails: the HOLDOUT FAILED with a NEGATIVE Sharpe -3.51 (ratio -0.535), and sensitivity FAILED with 2 cliffs on the core signal parameters tp_atr_mult ([email protected] -> [email protected]) and volume_mult ([email protected] -> [email protected]). This is the family's recurring signature: an encouraging, non-overfit walk-forward (PBO 0.25, is_overfitted=FALSE, all three OOS windows positive, avg OOS 6.56) and a positive bootstrap CI (sharpe_ci_low 1.30, PSR 0.99) sit on top of a deeply negative chronological holdout, because the resampled walk-forward and bootstrap both draw from the same bull-dominated 2017-2024 history while the time-ordered holdout exposes the recent decay (2026 annual -6.3%). The PSR 0.99 / DSR 0.02 split is the diagnostic tell that the headline is best-of-225 selection inflation, not edge. Not iterate (attempt 1 of 2): the failure is a bull-concentrated edge that goes negative in the recent holdout plus cliffs on the two defining parameters and a DSR far below the luck bar — not an under-explored robust region, so a second sweep re-selects another bull-fit config that fails the 2026 holdout. Not revise_hypothesis: the volume-confirmed Donchian breakout mechanism is already validated and landed on the liquid majors; ETH Daily is the thin, decaying arm of the same family's selectivity, not a proven mechanism stranded on a dead target. FAILURE PATTERN: long-only Daily volume-confirmed Donchian breakout on ETH optimizes to a flashy Sharpe (5.71) with an unusually clean robustness profile (PBO 0.25, non-overfit walk-forward, positive OOS windows, positive CI), yet still fails deflated Sharpe (0.02, far below a 10.69 luck bar), a NEGATIVE holdout (-3.51), and sensitivity cliffs on tp_atr_mult/volume_mult — the identical bull-concentrated, holdout-failing signature as the abandoned ADA (DSR 0.13, holdout -3.63), AVAX (DSR 0.06, holdout -3.78), and SOL-Daily (holdout -1.64) siblings. A clean, non-overfit walk-forward and a positive whole-sample CI do not rescue a breakout edge whose chronological holdout loses money; the time-ordered holdout is the gate that exposes the regime decay the resampled metrics mask.

Long-only ETHUSDT daily spot Volume-Confirmed Donchian Breakout (pure 3-filter mechanism, no regime overlays). Enters when price breaks above the prior 20-day Donchian high (measured in ATR units past the channel), confirmed by above-average volume and an uptrend gate (close > 50-day SMA). Exits on an ATR take-profit, ATR stop-loss (both snapshotted at entry), or a turtle-style Donchian low channel break. ATR-risk-based sizing capped at a notional fraction of equity.

Verification failed (Layer 2 — synthetic scenarios): Parameters used: ['risk_pct', 'atr_period', 'sma_period', 'sl_atr_mult', 'tp_atr_mult', 'volume_mult', 'min_notional', 'size_precision', 'entry_atr_buffer', 'volume_ma_period', 'max_notional_frac', 'donchian_exit_period', 'donchian_entry_period'] Check that __init__ sets all attributes from self.parameters.get(). - steady_uptrend: TypeError: EthDailyVolumeConfirmedDonchianBreakoutLong._mean() takes 1 positional argument but 2 were given (bar timestamp: 1735692720000) - steady_downtrend: TypeError: EthDailyVolumeConfirmedDonchianBreakoutLong._mean() takes 1 positional argument but 2 were given (bar timestamp: 1735692720000) - flat_ranging: TypeError: EthDailyVolumeConfirmedDonchianBreakoutLong._mean() takes 1 positional argument but 2 were given (bar timestamp: 1735692720000) - volatility_spike: TypeError: EthDailyVolumeConfirmedDonchianBreakoutLong._mean() takes 1 positional argument but 2 were given (bar timestamp: 1735692720000) - zero_volume: TypeError: EthDailyVolumeConfirmedDonchianBreakoutLong._mean() takes 1 positional argument but 2 were given (bar timestamp: 1735692720000) - price_gap: TypeError: EthDailyVolumeConfirmedDonchianBreakoutLong._mean() takes 1 positional argument but 2 were given (bar timestamp: 1735692720000)

Trade pattern implements the hypothesis: 58 long volume-confirmed breakout trades with asymmetric payoff (avg_win $6,845 > avg_loss $3,969, PF 1.61) — a valid trend-following profile

Clean, non-artifact metrics: kurtosis 1.42, skew 0.21, max_DD 15.8%, returns spread across multiple years (2020/2021/2023/2024 all ~13-21%) rather than one-bull-concentrated

Liquid-major arm of the family on a confirmed-working ETH Daily data path; single-instrument/single-timeframe avoids stale-leg failure mode

Sample (58 trades) clears the ~30-trade threshold for a walk-forward attempt

Trade frequency ~6.4/year is far below the hypothesis's claimed 15-25/year; split across 3 walk-forward windows the OOS windows risk being thin/zero-trade (the failure mode that sank the BNB and SOL-Daily siblings)

sharpe_ci_low is -0.51 — headline Sharpe 3.75 has a CI straddling 0 pre-optimization; edge not yet distinguishable from noise

2026 returns -11% — recent decay consistent with this family's repeated forward-failure signature on the alt arms

Walk-forward is non-overfit (is_overfitted=FALSE, PBO 0.2474) with all three OOS windows positive (avg OOS 6.56)

Optimized in-sample profile is clean (Sharpe 5.71, PF 1.99, max_DD 11.4%, sharpe_ci_low 1.30 positive, returns spread across 2017-2024 rather than one bull)

Liquid-major single-instrument design with no stale-leg risk

Failed deflated Sharpe: DSR 0.0208, is_significant=FALSE; optimized Sharpe 5.71 far below the expected-max luck bar 10.69 over 225 trials

HOLDOUT FAILED with a NEGATIVE Sharpe -3.51 (ratio -0.535) — the recent chronological window loses money

Sensitivity FAILED with 2 cliffs on core signal parameters tp_atr_mult ([email protected] -> [email protected]) and volume_mult ([email protected] -> [email protected])

Recent decay: 2026 annual -6.3%, consistent with the negative holdout — the edge is concentrated in earlier bull regimes

PSR 0.99 vs DSR 0.02 split confirms the headline Sharpe is best-of-225 selection inflation, not a generalizing edge

EthDailyVolumeConfirmedDonchianBreakoutLong applied the validated Volume-Confirmed Donchian breakout template to the ETH Daily gap with a clean single-instrument spot design, posting a 72.6% base return with genuine payoff asymmetry and earning an 'optimize' verdict. Optimization was unusually encouraging by resampled measures — a non-overfit walk-forward with PBO 0.25 and three positive OOS windows averaging 6.56 — but the forward gates still broke: the chronological holdout went deeply negative (-3.51), the deflated Sharpe sat at 0.02 against a 10.69 luck bar, and sensitivity cliffs appeared on two defining parameters. The PSR 0.99 / DSR 0.02 split confirmed the headline was best-of-225 selection inflation over a bull-concentrated 2017-2024 history that decays by 2026. On its second iteration the analyst abandoned it, citing the identical holdout-failing signature as the abandoned ADA, AVAX, and SOL-Daily siblings.

A clean, non-overfit walk-forward and a positive whole-sample bootstrap CI do not rescue a breakout edge whose time-ordered holdout loses money — resampled walk-forward and bootstrap both draw from the same bull-dominated history, while only the chronological holdout exposes the recent regime decay.

Despite a non-overfit walk-forward (is_overfitted=FALSE, PBO 0.25, all three OOS windows positive, avg OOS 6.56) and a positive optimized CI low (1.30), the analyst abandoned it because the chronological holdout FAILED with a negative Sharpe of -3.51 (ratio -0.535), the deflated Sharpe was 0.02 far below the 10.69 luck bar (is_significant=FALSE), and sensitivity FAILED with cliffs on the core tp_atr_mult and volume_mult parameters.

A long-only, single-instrument ETHUSDT daily spot strategy implementing the pure three-filter Volume-Confirmed Donchian Breakout template — a 20-day Donchian high breakout plus volume confirmation plus a 50-day SMA trend gate — deliberately stripped of regime overlays to keep trade count high for walk-forward statistical power.

The base backtest returned 72.6% total (CAGR 6.6%) over 58 volume-confirmed breakout trades with asymmetric payoff (avg win $6,845 vs avg loss $3,969, PF 1.61), max drawdown 15.8%, and Sharpe 3.75 — though its CI low of -0.51 straddled zero. Optimization produced a cleaner-looking profile (Sharpe 5.71, PF 1.99, max DD 11.4%, 52 trades, 65.4% win rate) with an unusually robust walk-forward.