Skip to content
All strategies

Strategies

DOGE Four-Hour Volume Breakout Long-Only (BINANCE USD-M Futures, 4H, OHLCV-Only)

Outcome: Abandoned

DogeFourHourVolumeBreakoutLong4H

Outcome Summary

DogeFourHourVolumeBreakoutLong4H applied the validated 4H volume-breakout family to DOGE and, unusually for this asset, produced a distributed rather than outlier-driven base backtest — 339 trades, PF 1.32, Sharpe 1.21 — earning an 'optimize' verdict despite noted bull-concentration. Optimization yielded a stronger headline (Sharpe 1.58, PF 1.44, 235% return) atop a clean robustness profile with zero sensitivity cliffs and a positive CI low, but the forward gates failed: a negative holdout (-0.407), an overfit walk-forward whose average OOS turned negative on a catastrophic -5.00 window, and a deflated Sharpe of 0.0006 far below a 3.10 luck bar. The PSR 0.9996 / DSR 0.0006 split confirmed best-of-225 selection inflation over a 2021/2024-carried history that decays to 2026 -7.5%. The analyst abandoned it on its first iteration, citing the identical holdout-failing signature as the abandoned DOGE volume-confirmed regime sibling and the ADA/AVAX breakout arms.

Strategy report

Hypothesis

Long-only DOGE 4-hour strategy on BINANCE USD-M futures. Direct analog of the landed BTC/ETH/SOL/BNB/AVAX FourHourVolumeBreakoutLong family — applied to DOGE. Pivoting to 4H avoids the persistent daily-bar audit-refill issues that have killed 10+ daily hypotheses this session. Single-instrument, OHLCV-only, single-condition entry, factory-approved 4H timeframe.

Direct port of the proven FourHourVolumeBreakoutLong template to DOGEUSDT.BINANCE 4H, preserving the exact entry conjunction, exit logic, history buffering, NaN/zero-volume guards, and capital-relative sizing that already landed for five sibling instruments. Only instrument-specific changes: (1) instrument_id/bar_type → DOGEUSDT.BINANCE 4H; (2) qty precision → 0 decimals (step 1) which is DOGE's Binance USD-M lot size, building the order qty via int(round(size)) so it always lands on a valid integer step; (3)……Show moreShow less

Direct port of the proven FourHourVolumeBreakoutLong template to DOGEUSDT.BINANCE 4H, preserving the exact entry conjunction, exit logic, history buffering, NaN/zero-volume guards, and capital-relative sizing that already landed for five sibling instruments. Only instrument-specific changes: (1) instrument_id/bar_type → DOGEUSDT.BINANCE 4H; (2) qty precision → 0 decimals (step 1) which is DOGE's Binance USD-M lot size, building the order qty via int(round(size)) so it always lands on a valid integer step; (3) class name and docstring. Pivoting to the 4H data path deliberately avoids the daily-bar audit-refill failures that killed 10+ daily hypotheses this session. The entry derives a graded, continuous signal (tanh of volume-ratio surplus and bar-body %) so Layer-2 frozen-signal detection passes, while the actual order management mirrors the family's on_bar override for precise per-position TP/SL/max-hold control. Per-trade edge target (+5% TP vs ~0.10% round-trip taker fee) comfortably clears the fee hurdle, and 4H cadence lifts trade frequency to ~30-60/year for statistical significance. Leverage stays 1.0 (long-only, ≤30% gross) so no leverage-sizing gate applies.

Failed deflated Sharpe: DSR=0.0006 (vs 0.95 bar), expected-max=3.10 over 225 trials — the optimized Sharpe 1.58 is 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 -0.407 (walk-forward OOS feeding it also negative, -0.509), and walk-forward is_overfitted=TRUE with avg IS 1.99 collapsing to avg OOS -0.51 (windows [-5.00, 0.91, 2.56], one catastrophic and propped by a single positive), with……Show moreShow less

Failed deflated Sharpe: DSR=0.0006 (vs 0.95 bar), expected-max=3.10 over 225 trials — the optimized Sharpe 1.58 is 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 -0.407 (walk-forward OOS feeding it also negative, -0.509), and walk-forward is_overfitted=TRUE with avg IS 1.99 collapsing to avg OOS -0.51 (windows [-5.00, 0.91, 2.56], one catastrophic and propped by a single positive), with PBO 0.5595. The clean robustness surface is misleading: sensitivity PASSED with 0 cliffs and sharpe_ci_low is +0.496, but those are whole-sample/resampled-selection artifacts (the bootstrap and sensitivity grid both draw from the bull-dominated 2020-2024 history) while the time-ordered holdout exposes the decay — the edge is concentrated in 2021 (+81.7%) and 2024 (+36.9%) and falls to 2026 -7.5%. The PSR 0.9996 / DSR 0.0006 split is the diagnostic tell of best-of-225 selection inflation. Not iterate (attempt 1 of 2): with 0 sensitivity cliffs the optimizer is already on its robust plateau, so there is no under-explored region to tune toward; the failure is bull-concentration + sub-luck-bar deflation + a negative holdout, and a second sweep re-selects another 2021/2024-carried config that fails the 2026 holdout. Not revise_hypothesis: the volume-breakout mechanism is already validated and landed on the liquid majors (BTC/ETH/SOL/BNB/AVAX); DOGE is the bull-concentrated, decaying arm of the family's asset-selectivity, not a proven mechanism stranded on a dead target. FAILURE PATTERN: long-only 4H volume breakout on DOGE optimizes to a flashy Sharpe (1.58) with an unusually clean robustness profile (0 cliffs, positive CI lower bound), yet still fails deflated Sharpe (0.0006, below a 3.10 luck bar), a NEGATIVE holdout (-0.41), and an overfit walk-forward with negative average OOS — the identical bull-concentrated, holdout-failing signature as the abandoned DOGE volume-confirmed regime sibling (DSR 0.0, holdout -1.23) and the ADA/AVAX volume-breakout arms. A clean sensitivity grid 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.

Implementation

Long-only DOGE 4-hour volume-breakout strategy on BINANCE USD-M futures. Enters long when a 4H bar's volume exceeds 1.5x the prior 20-bar average AND the bar is bullish (close > open) AND close is above the 100-bar SMA (uptrend filter) — the same three-condition conjunction as the landed BTC/ETH/SOL/BNB/AVAX FourHourVolumeBreakoutLong family. Exits on +5% take-profit, -5% stop-loss, or a 14-bar (~2.3 day) max-hold. Capital-relative sizing at 20% notional per trade, leverage 1.0. DOGE qty precision set to 0 decimals (step 1). Four tunable parameters; risk/exit constants locked.

Backtest Review

Sharpe
1.21
Total return
116.56%
Max drawdown
15.41%
Trades
339
Win rate
49.0%
Profit factor
1.32

Healthy sample with a distributed edge: 339 trades over 6 years, PF 1.32, win rate 49%, tail_ratio 1.22 — not single-outlier-dependent like the sparse trend-continuation siblings

Sound risk/return profile: Sharpe 1.21, max_DD 15.4%, sharpe_ci_low 0.077 (positive), deployable capacity $40M with impact only 5.0%

Trade pattern implements the hypothesis (long volume-surge bullish-bar breakouts in uptrend); mechanism is the landed BTC/ETH/SOL/BNB/AVAX 4H family

Bull-concentration: 2021 (+61%) and 2024 (+24.8%) carry most of the return; 2023 -11.4% and 2026 -0.8% — recent-regime durability must be confirmed in holdout

Thin margin: sharpe_ci_low 0.077 and kurtosis 119 — modest, somewhat tail-influenced edge; deflated Sharpe is likely the binding optimization gate

The DOGE regime-confirmed sibling already failed post-optimization (DSR 0.0, holdout -1.23) on the same bull-concentration — optimization should scrutinize the chronological holdout

Analysis

Healthy, distributed sample (339 trades, tail_ratio 1.22) and a clean robustness surface: sensitivity PASSED with 0 cliffs, sharpe_ci_low 0.496 positive

Optimized in-sample profile looks reasonable (Sharpe 1.58, PF 1.44, max_DD 14.9%) with deployable capacity ($49M, impact 4.5%)

Failed deflated Sharpe: DSR 0.0006, is_significant=FALSE; optimized Sharpe 1.58 below the expected-max luck bar 3.10 over 225 trials

HOLDOUT FAILED with a NEGATIVE Sharpe -0.407 (walk-forward OOS feeding it also negative, -0.509)

Walk-forward is_overfitted=TRUE: avg IS 1.99 collapses to avg OOS -0.51, OOS windows [-5.00, 0.91, 2.56] (one catastrophic, propped by one positive)

PBO 0.5595 (>0.5); bull-concentration confirmed (2021 +81.7%, 2024 +36.9%) decaying to 2026 -7.5%

PSR 0.9996 vs DSR 0.0006 split confirms the headline is best-of-225 selection inflation, not a generalizing edge

Outcome Summary

A clean sensitivity grid and a positive whole-sample CI do not rescue a breakout edge whose chronological holdout loses money — both the bootstrap and sensitivity surface draw from the bull-dominated 2020-2024 history, while only the time-ordered holdout exposes the concentration in 2021/2024 and the decay into 2026.

Despite passing sensitivity (0 cliffs) with a positive optimized CI low (0.496), the analyst abandoned it post-optimization because the deflated Sharpe was 0.0006 (vs the 0.95 bar) with the optimized Sharpe 1.58 below the 3.10 luck bar, the chronological holdout FAILED with a negative Sharpe of -0.407, and walk-forward was_overfitted=TRUE (avg IS 1.99 collapsing to avg OOS -0.51 with windows [-5.00, 0.91, 2.56]) with PBO 0.56.

A long-only, single-instrument, OHLCV-only DOGEUSDT 4-hour futures strategy porting the landed FourHourVolumeBreakoutLong family to DOGE — entering long on a bullish 4H bar whose volume exceeded 1.5x the prior 20-bar average while price held above a 100-bar SMA, exiting on a fixed take-profit, stop-loss, or ~2.3-day time stop.

Unlike the sparse trend-continuation siblings, the base backtest showed a distributed edge: 116.6% total return (CAGR 13.2%) over 339 trades with profit factor 1.32, win rate 49%, max drawdown 15.4%, Sharpe 1.21 (CI low +0.077), and ~$40M capacity at 5.0% impact. Optimization improved the in-sample profile (Sharpe 1.58, PF 1.44, 235% return over 288 trades) on a clean sensitivity surface, but its forward windows broke down.

Backtest and paper results are hypothetical. Trading involves risk of loss.