EthDailyCounterRallyShortDowntrend

ETH Daily Counter-Rally Short in Confirmed Downtrend (Bear-Flag Bounce Failure)

A SHORT-ONLY strategy on ETHUSDT.BINANCE perpetual futures (Binance USD-M) using DAILY bars. Entry on 2 consecutive UP closes WHILE the instrument is in a confirmed long-term DOWNTREND (close < 200-SMA, 50-SMA < 200-SMA). This is the EMPIRICAL INVERSE of the just-falsified XRP pullback strategy: the XRP failure proved that bounces in uptrends do NOT hold and prices keep falling. The contrapositive is that BOUNCES IN DOWNTRENDS also do not hold — rallies during confirmed bear regimes are failed bounces that resume the underlying downtrend. This bear-flag-bounce-failure pattern is documented in price-action literature (Brooks 2009 'Trading Price Action Trends' calls it a 'major-trend-reversal-rejection' setup) and historically works in trending markets where counter-trend rallies fail to reclaim the broken support. ETH is selected because: (1) the pipeline has TWO ETH LONG strategies and ZERO ETH shorts — adding a short directly diversifies within instrument; (2) ETH has clean Binance USD-M data since 2019 with no deprecation risk; (3) ETH had multiple multi-month downtrend regimes (2022 bear, March 2024 pullback, 2025 chop) that produce many qualifying setups; (4) ETH precision is in the map (proven by 2 existing pipeline strategies). Structurally distinct from every pipeline strategy and failed mechanism: NOT a long-only strategy (only the BTC death-cross is short, but it uses SMA-crossover entry not consecutive-up-day count); NOT a pullback long (the just-falsified class); NOT supp-data dependent; NOT calendar; NOT accumulation; NOT cross-asset. The mechanism explicitly leverages the session's most-decisive empirical finding (XRP pullback −7.15 Sharpe = prices keep moving with regime) by inverting the trade direction. Critical advantages over the failed XRP setup: (1) trades WITH the documented regime persistence (short in downtrend), not against it; (2) ETH not XRP — different microstructure to avoid pattern-specific overfitting; (3) shorter holding period (7 days max vs XRP's reclaim-wait) so trades resolve quickly into the dominant trend direction.

Iteration-2 minimal fix for the Layer-2 TypeError (_sma() takes 2 positional arguments but 3 were given), with no earlier-layer regression. Root cause: _sma 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 helper was called with 3 args and raised. The fix moves _sma to a module-level function (module functions are never bound to an instance, so no self can be injected) and updates the two call sites to call it directly. Nothing else changed: imports stay clean (Layer 1 green), and the regime/consecutive-up-close/signal/entry/exit/sizing logic is byte-identical (Layer 2 logic and Layer 3 behavior preserved). Venue remains BINANCE USD-M futures because the strategy is short-only and needs a margin account; leverage 1.0 (not used for amplification, so the leverage_set_but_unused gate does not trigger).

Bear-regime-concentrated short that cannot be validated out-of-sample — exactly the risk flagged at pre-optimization. Optimization degraded the base (Sharpe 0.31 -> optimized -0.64, total_return -8.3%, PF 0.79, expectancy -$184/trade) by driving sma_slow 200->278 and up_closes 2->3, thinning the sample to 41 trades. Every gate fails: deflated_sharpe 0.0 (is_significant FALSE) with the optimized Sharpe -0.64 far below the expected-max luck bar 4.04, PBO 0.7289, sharpe_ci_low -2.47 (CI straddles 0); walk-forward is_overfitted=TRUE with avg IS 1.93 collapsing to avg OOS 0.0 and ALL three OOS windows producing ZERO trades; the holdout FAILED with zero trades (holdout_sharpe 0.0); and sensitivity FAILED with 2 cliffs (sma_slow, stop_loss_pct). The edge is entirely 2022-bear-concentrated (+10.1% in 2022, then every year negative: 2023 -9.0%, 2024 -2.2%, 2025 -3.3%, 2026 -2.4%). Not promote: a negative optimized Sharpe + DSR 0.0 + is_overfitted with zero-trade OOS + a zero-trade holdout + PBO 0.73 + cliffs are each disqualifying. Not iterate (attempt 1 of 2): the failure is structural regime-rarity -- a short that only fires in confirmed downtrends produces no trades in non-bear OOS/holdout windows, and no parameter change manufactures bear-regime setups where none exist; the base 76-trade config is the densest version and is already a sub-luck-bar, 2022-carried edge, so a second sweep re-finds another 2022-concentrated config with empty recent windows. Not revise_hypothesis: the bear-rally-fade short has no promoted sibling and is intrinsically rare on a structurally up-trending major. FAILURE PATTERN: short-only counter-trend (bear-rally-fade) strategies on a crypto major concentrate their entire edge in one sustained bear (2022) and cannot be validated out-of-sample -- the regime filter starves the walk-forward OOS and holdout windows of trades (zero-trade OOS/holdout), so best-of-225 selection drives the config to a 2022-fit that fails deflated Sharpe (0.0), is_overfitted=TRUE, with PBO 0.73; a regime-gated short whose setups exist only ~20% of the time is structurally unvalidatable on a single asset.

Short-only counter-rally fade on ETHUSDT.BINANCE USD-M futures, DAILY bars, pure OHLCV. Shorts a failed bounce (>=2 consecutive up-closes) while the long-term regime is bearish (close < 200-SMA and 50-SMA < 200-SMA). Exits on an 8% take-profit, 6% hard stop, regime break (close back above the 200-SMA), or a 7-day time stop. The signal is the continuous close-vs-200-SMA distance (negative in a downtrend, varies every bar); the discrete short trigger gates entry. Sized at 40% equity notional, leverage 1.0.

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

Implemented correctly and trades as specified: 76 short entries (0 long), 0 dropped, regime-gated to confirmed 50/200-SMA downtrends — sound bear-rally-fade mechanism that trades WITH regime persistence

Net-positive with per-trade above costs (~0.3%), decent capacity (~$17M); fills a genuine portfolio gap (ETH short)

Useful hedge profile: beta -0.056, benchmark_correlation -0.39 — earns +27% in 2022 when long strategies bleed; diversification value beyond standalone Sharpe

Thin and bear-concentrated: PF 1.091, Sharpe 0.31 (CI -1.19 to 1.69), edge carried by 2022 (+27.4%) and 2026 (+10.0%); negative in 2023/2024/2025 with rolling Sharpe to ~-4.4

Sparse: 76 trades (~12/yr) — walk-forward windows are thin and the holdout depends on whether the recent window contains a tradeable bear regime

max_drawdown 24.2% (CI to 44.5%), max_consecutive_losses 7, kurtosis 16.2 — some outlier dependence

Marginal standalone metrics: the DSR will test whether a 0.31-Sharpe, 76-trade, regime-concentrated edge is significant

Implemented correctly and trades as specified (short-only, 76->41 entries, 0 dropped); useful negative correlation (beta -0.056, benchmark_correlation -0.39) and 2022 bear performance (+10-27%)

Optimization degraded it to a net loser: base Sharpe 0.31 -> optimized -0.64, total_return -8.3%, profit_factor 0.79, expectancy -$184/trade

Failed deflated Sharpe: DSR=0.0 with the optimized Sharpe -0.64 far below the expected-max luck bar 4.04; PBO 0.7289; sharpe_ci_low -2.47 (CI straddles 0)

Walk-forward is_overfitted=TRUE: avg IS 1.93 -> avg OOS 0.0 with ALL three OOS windows producing ZERO trades; holdout FAILED with zero trades (holdout_sharpe 0.0)

Sensitivity FAILED with 2 cliffs (sma_slow, stop_loss_pct)

Edge is 2022-bear-concentrated (every year after 2022 negative); the short only fires in sustained downtrends, so recent/holdout windows have no setups

This strategy inverted a freshly-falsified pullback-long into a bear-rally-fade short on ETH, filling a genuine portfolio gap and trading with documented regime persistence. The base looked defensible as a hedge — net-positive, negatively correlated, +27% in the 2022 bear — and earned an 'optimize' verdict despite flagged sparsity and bear-concentration. Optimization confirmed the worry: it degraded the edge to a negative Sharpe and thinned the sample, and every forward gate failed — deflated Sharpe 0.0, PBO 0.73, an overfit walk-forward with all three OOS windows and the holdout producing zero trades, plus two sensitivity cliffs. The analyst abandoned it on iteration 2 as structurally unvalidatable: a short whose setups exist only in sustained downtrends (~20% of the time) leaves recent windows empty, so its entire edge is a 2022-concentrated fit with no robust region to tune toward.

A regime-gated short-only counter-trend strategy on a structurally up-trending major concentrates its entire edge in one sustained bear (2022) and is intrinsically unvalidatable out-of-sample — the regime filter starves walk-forward OOS and holdout windows of trades, so best-of-N selection just refits the lone bear and fails deflated Sharpe; no parameter change manufactures bear setups where none exist.

The post-optimization analyst abandoned it on every gate: deflated Sharpe 0.0 (optimized Sharpe -0.64 far below the 4.04 luck bar), PBO 0.7289, walk-forward is_overfitted=True with all three OOS windows producing zero trades, a failed zero-trade holdout, and two sensitivity cliffs — with iteration ruled out because a short that only fires in confirmed downtrends is structurally starved of trades in non-bear OOS/holdout windows.

A short-only daily counter-rally fade on ETHUSDT perpetual futures (pure OHLCV) — the empirical inverse of a falsified pullback-long — shorting 2 consecutive up-closes (a failed bounce) only while the long-term regime is bearish (close < 200-SMA and 50-SMA < 200-SMA), exiting on an 8% take-profit, 6% stop, regime break, or 7-day time stop.

The base backtest was thin but net-positive and a useful hedge — 76 short entries (none long, none dropped), +7.1% total, Sharpe 0.31 (CI -1.19 to 1.69), profit factor 1.091, 24.2% max drawdown, beta -0.056 and benchmark correlation -0.39 — but bear-concentrated (carried by 2022 +27.4% and negative in 2023/2024/2025); optimization degraded it to a net loser (Sharpe -0.64, -8.3% total, PF 0.79, expectancy -$184/trade, thinned to 41 trades).