EthDailyHigherHighHigherLowDowTrendConfirmLong

ETH Daily Higher-High-Higher-Low Dow Theory Trend Confirmation Long with 200-SMA Bull Regime Filter

A long-only single-instrument MULTI-BAR DOW THEORY CONFIRMATION strategy on ETHUSDT perpetual futures using daily bars and OHLCV-only data, with a 200-SMA bull regime filter. The trigger is the classical 'Higher High + Higher Low' (HH/HL) trend-confirmation pattern that traces to Charles Dow's editorials (1900s), was codified by Robert Rhea in 'The Dow Theory' (1932), and remains a foundational tool in classical technical analysis. The mechanism requires the CURRENT daily bar to simultaneously satisfy TWO positional dominance conditions over a 5-bar lookback window: (a) current HIGH must exceed all 5 prior daily highs (positional dominance on the upside), AND (b) current LOW must exceed all 5 prior daily lows (positional dominance on the downside, indicating no deep retracement during the bar). When both conditions hold simultaneously on the same bar, Dow Theory considers the structural uptrend to be confirmed for continuation: buyers have demonstrated control across the entire daily range. This mechanism is structurally NOVEL within the portfolio and complementary to every existing strategy: candlestick patterns (Hammer, Engulfing, Morning Star, Three White Soldiers) use BODY/SHADOW PROPORTIONAL geometry within 1-3 bars; pivot patterns (Pivot Low) use rank-based detection of local extrema; this HH/HL pattern uses PURE POSITIONAL DOMINANCE over a 5-bar window — both bar EXTREMES (high AND low) must out-rank the entire prior window. No existing strategy uses dual positional dominance. The 5-bar lookback distinguishes this from breakout strategies (which only require high > prior high) — the additional low > prior low constraint dramatically reduces false signals in volatile chop. ETH is chosen specifically because (a) ETH is now the most UNDERUTILIZED asset in the portfolio (only EthDailyGoldenCrossMomentumLong remains after Bullish Engulfing failure — 1 strategy vs 2-3 for every other tracked asset), (b) ETH daily futures data path is empirically reliable (used by Golden Cross + multiple regime filters), (c) ETH exhibits clean Dow-Theory trending behavior driven by deep institutional flows (ETF rebalancing, staking redemption cycles produce structurally clean HH/HL sequences), (d) ETH's lower volatility relative to alts produces less HH/HL noise, making the signal cleaner. Only 3 explicit parameters (regime SMA period, lookback window, stop-loss pct) — minimum-parameter pattern strategy.

Iteration-2 minimal fix for the Layer-2 TypeError (_sma() takes 2 positional arguments but 3 were given). Root cause is identical to the open-interest strategy's fix: the SMA helper was a @staticmethod, but the Layer-2 _StrategyProxy rebinds class callables onto a stand-in instance, which discards the staticmethod descriptor and passes the proxy instance as an injected first positional argument — so the 2-param _sma is called with 3 args and raises. The fix moves _sma out of the class to a module-level function; module functions are never bound to an instance, so no self can be injected, and calculate_signal now calls _sma(self._close, self.regime_sma) directly. Nothing else changed: imports stay clean (Layer 1 green), and the dual-dominance HH+HL signal, regime filter, entry/exit, and sizing logic the harness exercised are byte-identical, so no earlier-layer regression. Venue remains BINANCE USD-M futures, leverage 1.0 (no amplification, so the leverage_set_but_unused gate does not trigger).

Front-loaded edge that dies out-of-sample — every decisive generalization gate fails in concert, exactly as flagged at pre-optimization. The strategy is implemented correctly and trades as specified (long-only dual positional-dominance HH/HL, 104->91 ETH-daily entries), but the multiple-testing and forward gates reject it: deflated_sharpe 0.0 (is_significant FALSE) with the optimized Sharpe 0.22 far below the expected-max luck bar 5.06 over 225 trials, PBO 0.6548 (>0.5), sharpe_ci_low -1.79 (CI straddles 0); walk-forward is_overfitted=TRUE with avg IS 0.83 collapsing to avg OOS -5.74 (OOS windows [-7.87, 0.0, -9.35], one zero-trade); and the holdout FAILED (holdout_sharpe 0.0, ratio 0). Sensitivity also flags the CORE signal param lookback as a cliff (Sharpe 0.39->0.065 at 5->6). Optimization degraded the strategy (Sharpe 0.39->0.22, return 42.6%->15%), and the optimized annual returns are carried almost entirely by 2024 (+10.6%) with 2021/2023/2025 negative — a bull-regime-concentrated signal that does not generalize. Not iterate (attempt 1 of 2): there is no robust parameter region to tune toward when DSR=0.0, PBO=0.65, avg OOS=-5.74, the holdout fails, and the core signal param is a cliff — a second sweep re-selects another sub-luck-bar artifact. Not revise_hypothesis: this is overfitting + regime-concentrated decay, not a promoted mechanism stranded on a dead target (dual-dominance HH/HL has no promoted sibling to redirect, and reframing onto another major re-hits the same weak bull-only edge). FAILURE PATTERN: single-asset positional-dominance (HH/HL Dow-Theory) long-only on a crypto major produces a 2024-concentrated edge that fails deflated Sharpe (0.0, vs a 5.06 expected-max luck bar), is_overfitted=TRUE with deeply negative average OOS (-5.74), and a failed holdout after best-of-225 selection — pure-price trend-confirmation patterns on majors do not generalize beyond bull regimes and cannot be validated out-of-sample.

Long-only Dow-Theory Higher-High+Higher-Low trend-confirmation strategy on ETHUSDT.BINANCE USD-M futures, daily bars, pure OHLCV single feed. Enters LONG when the current daily bar's high exceeds all 5 prior highs AND its low exceeds all 5 prior lows (dual positional dominance) inside a bull regime (close > 200-SMA). Exits on an 8% hard stop, a regime break (close < 200-SMA), a structure break (current low < prior low), or a 20-bar time stop. The signal is a continuous dual-dominance score (min of high- and low-margins) recomputed every bar; the discrete HH+HL+regime condition gates entry. Sized at 95% equity notional, leverage 1.0.

Verification failed (Layer 2 — synthetic scenarios): Parameters used: ['lookback', 'regime_sma', 'min_notional', 'position_pct', 'max_hold_bars', 'stop_loss_pct'] Check that __init__ sets all attributes from self.parameters.get(). - steady_uptrend: TypeError: EthDailyHigherHighHigherLowDowTrendConfirmLong._sma() takes 2 positional arguments but 3 were given (bar timestamp: 1735701660000) - steady_downtrend: TypeError: EthDailyHigherHighHigherLowDowTrendConfirmLong._sma() takes 2 positional arguments but 3 were given (bar timestamp: 1735701660000) - flat_ranging: TypeError: EthDailyHigherHighHigherLowDowTrendConfirmLong._sma() takes 2 positional arguments but 3 were given (bar timestamp: 1735701660000) - volatility_spike: TypeError: EthDailyHigherHighHigherLowDowTrendConfirmLong._sma() takes 2 positional arguments but 3 were given (bar timestamp: 1735701660000) - zero_volume: TypeError: EthDailyHigherHighHigherLowDowTrendConfirmLong._sma() takes 2 positional arguments but 3 were given (bar timestamp: 1735701660000) - price_gap: TypeError: EthDailyHigherHighHigherLowDowTrendConfirmLong._sma() takes 2 positional arguments but 3 were given (bar timestamp: 1735701660000)

104 long-only entries over ~6.5 years (~16/yr) — healthy sample for sensitivity/walk-forward; not too sparse to optimize

Mechanism implemented correctly: entry_diag shows 104 signaled -> 104 submitted, 0 dropped (size or min-notional); all entries long, matching the Dow-Theory HH/HL long-only premise

Selective exposure (14.3%) consistent with a trend-confirmation pattern; positive total return (+42.6%) and PF > 1, so the signal is functioning and not degenerate

Only 3 primary params (regime_sma, lookback, stop_loss_pct) — low overfitting surface for optimization

Edge is thin: profit_factor 1.066, Sharpe 0.394, win_rate 41% — barely above costs

Returns heavily concentrated in 2024 (+47.4%); negative in 2021 (-18.8%) and 2023 (-9.4%) — regime-dependent, a forward-validation risk

High modeled market impact (impact_cost_pct 16.2%) with low capacity (~$3.8M); avg_position_pct ~116% with leverage 1.0 — impact eats a large share of gross edge

max_drawdown 37% with a 1466-day drawdown duration; return_kurtosis 27 and largest_win ($27.6k) suggest fat-tail dependence

Implemented correctly and trades as specified (104 long HH/HL entries initially, 91 after optimization, 0 dropped) — no code defect

Sensitivity grid is mostly smooth on regime_sma/position_pct/stop_loss (only lookback flagged as a cliff)

Failed deflated Sharpe: DSR=0.0 (bar 0.95), optimized Sharpe 0.22 far below the expected-max luck bar 5.06 over 225 trials — indistinguishable from best-of-N selection noise

is_significant FALSE; sharpe_ci_low -1.79 straddles 0; PBO 0.6548 (>0.5 ⇒ more likely overfit than not); PSR 0.597

Walk-forward is_overfitted=TRUE: avg IS 0.83 collapses to avg OOS -5.74, OOS windows [-7.87, 0.0, -9.35] (one zero-trade window)

Holdout FAILED: holdout_sharpe 0.0, ratio 0 against a deeply negative WF-OOS denominator

Core signal parameter lookback is a sensitivity cliff (Sharpe 0.39 -> 0.065 moving 5 -> 6)

Optimization degraded the strategy (Sharpe 0.39 -> 0.22, total return 42.6% -> 15%); edge concentrated in 2024 (+10.6%), negative in 3 of 4 years (2021/2023/2025)

High impact cost (16-21%) and small capacity (~$2.2-3.8M) — fee/impact fragile even if an edge existed

This strategy added a structurally novel dual-positional-dominance mechanism — Dow Theory higher-high-plus-higher-low confirmation — to underutilized ETH daily, gated by a 200-SMA bull filter. It earned an 'optimize' verdict on a functioning but thin edge (104 trades, Sharpe 0.39, +42.6%), though flagged as 2024-concentrated with high impact cost. Optimization confirmed the worry: deflated Sharpe 0.0 against a 5.06 luck bar, walk-forward overfitted with average OOS Sharpe -5.74, a failed holdout, a core-parameter cliff, and a degraded optimized result. The analyst abandoned it at the analyzing stage on its second iteration as a front-loaded edge that dies out-of-sample — no robust region to tune and no promoted sibling to redirect to — so it stopped before risk review.

A pure-price trend-confirmation pattern (HH/HL dual dominance) on a crypto major produces a thin, bull-regime-concentrated edge that does not generalize — a +42.6% base carried almost entirely by 2024 collapses under deflation (DSR 0.0) with deeply negative OOS and a failed holdout, and optimization made it worse rather than finding a robust region.

It cleared the backtest-review gate with an 'optimize' verdict, but the post-optimization analyst abandoned it: deflated Sharpe 0.0 (<0.95) with the optimized Sharpe 0.22 far below the 5.06 expected-max over 225 trials, is_significant false, PBO 0.65, a CI straddling zero, walk-forward overfitted (avg IS 0.83 → avg OOS -5.74, windows [-7.87, 0.0, -9.35]), a failed holdout, and the core lookback parameter flagged as a sensitivity cliff.

A long-only Dow Theory trend-confirmation strategy on ETHUSDT daily Binance futures (1.0x leverage, 0.95 position) requiring dual positional dominance — the current bar's high above all 5 prior highs AND its low above all 5 prior lows (higher-high + higher-low) — inside a 200-SMA bull regime, with structure-break, regime, stop, and time exits, on a minimal 3-parameter design.

The base backtest showed a thin edge — 104 trades, +42.6% total, Sharpe 0.39, profit factor 1.066, win rate 41%, a 37% max drawdown — heavily concentrated in 2024 (+47.4%) and negative in 2021 (-18.8%) and 2023 (-9.4%), with high modeled impact (16.2%) and ~$3.8M capacity; optimization degraded it further (Sharpe 0.22, +15%).