DogeFourHourTripleFilterTrendContinuationLong

DOGE 4H Triple-Filter Trend Continuation Long

A minimal long-only trend-following strategy on DOGEUSDT perpetual futures using 4-hour bars. Uses ONLY OHLCV data (no supplementary feeds) to keep verification simple and deterministic. The strategy enters long when three independent trend confirmations align: (1) price above the 50-period SMA, (2) the 50-SMA itself is sloping upward over the last 10 bars (long-term trend up), and (3) the most recent 4H close is in the upper third of its 20-bar Donchian range (recent strength). Exit uses a chandelier-style ATR trailing stop (3x ATR(14) below the highest close since entry) plus a hard time stop after 30 bars (~5 days). DOGE is chosen because (a) it has chronic high realized volatility on 4H (mean ATR/price ~2-3%), giving per-trade moves that comfortably exceed the 0.10% Binance futures round-trip cost, (b) it exhibits strong momentum-clustering during retail-driven hype waves, and (c) it is highly liquid with $1B+ daily futures volume. The intent is deliberately simple — three classic filters, one trailing stop, one time stop — to avoid both overfitting and verification hangs caused by complex multi-data-source strategies.

Implements the hypothesis exactly with the standard single-instrument four-hook pattern. calculate_signal computes all three filters every bar and returns a continuous graded value (close-vs-SMA50 distance in %) so Layer-2 frozen-signal detection passes, while the binary entry decision is thresholded in should_enter against the three stored filter flags (regime: close>SMA50; slope: SMA50 rising over slope_lookback via an offset SMA; strength: close in the upper upper_frac of the 20-bar Donchian range). should_exit implements the chandelier ATR trailing stop using the highest close since entry (tracked in calculate_signal against live position state, reset when flat) plus a 30-bar time stop. position_size is capital-relative and ATR-risk-based: risk 2% of equity over a 3x-ATR stop, capped at 50% notional so a single DOGE position never exceeds half of equity (leverage 1.0). min_bars_required = max(sma+slope, donchian, atr)+2 guarantees all indicators are valid before trading. Venue kept on BINANCE USD-M futures per the hypothesis: although the strategy is long-only with no leverage, futures round-trip cost (~0.10%) is actually cheaper than spot (~0.20%) and the hypothesis is explicitly built around DOGE perpetual-futures liquidity; DOGE's ~2-3% 4H ATR/price gives per-trade moves well above that cost. size_precision 0 (DOGE trades in whole-coin lot steps); make_qty refines to the instrument's exact precision.

Do not optimize — the expected outcome is a known dead end. The base has no edge before selection (Sharpe 0.95 with sharpe_ci_low 0.041, CI essentially touching zero), and the return is bull-concentrated and fat-tail-driven (return_kurtosis 164, skew 10.4; +219% carried by a handful of pump days such as 2021-04-15 +42.5% and 2024-03-01 +20.5%, plus two bull years 2021 +91% and 2024 +43%, with 2020/2023/2025 flat-to-negative). Best-of-225 optimization will inflate this marginal Sharpe and deflated Sharpe will then correct it back below the DOGE-4H luck bar (~3-5.8 in the sibling runs), while the time-ordered holdout will expose the bull-concentration. This is the third DOGE 4H long-only bull-concentrated trend/breakout strategy this session with the identical signature; the two siblings already failed post-optimization decisively — DogeFourHourVolumeBreakoutLong4H (DSR 0.0006, holdout -0.41) and the volume-confirmed-regime version (DSR 0.0, holdout -1.23) — both passing in-sample/sensitivity and dying in the chronological holdout. The triple-filter trend-continuation mechanism is generic and already validated on liquid majors; DOGE 4H is the bull-concentrated, holdout-failing arm of this family. Spending 2 hours of optimization to re-derive the same abandon is not justified; the idea should be abandoned rather than optimized or iterated (no parameter change converts a 2-bull-year, fat-tail-carried edge into a forward-robust one).

Long-only DOGEUSDT 4H trend-continuation strategy using only OHLCV. Enters long when three independent filters align: (1) close above the 50-SMA, (2) the 50-SMA is sloping up over the last 10 bars, and (3) close sits in the upper third of its 20-bar Donchian range. Exits on a chandelier ATR trailing stop (3x ATR(14) below the highest close since entry) or a 30-bar (~5 day) time stop. ATR-risk position sizing at 2% equity risk per trade, capped at 50% notional. Binance USD-M futures, leverage 1.0.

Well-formed and well-populated: 185 trades over 6 years, no correctness bug, deployable capacity ($54.7M, impact 4.3%), max_DD 17.5%

Simple, sensible 3-filter trend-continuation design

No edge before selection: Sharpe 0.95 with sharpe_ci_low 0.041 (CI essentially touches 0); best-of-225 optimization will inflate it and deflated Sharpe will correct it back below the luck bar

Bull-concentration + fat tails: return_kurtosis 164, skew 10.4; +219% carried by a few pump days (2021-04-15 +42.5%, 2024-03-01 +20.5%) and two bull years (2021 +91%, 2024 +43%), with 2020/2023/2025 flat-to-negative

Identical asset/timeframe/family signature to two DOGE 4H trend siblings already abandoned post-optimization this session (volume breakout: DSR 0.0006, holdout -0.41; regime-confirmed: DSR 0.0, holdout -1.23) — both bull-concentrated and holdout-failing

DogeFourHourTripleFilterTrendContinuationLong was a deliberately simple three-filter trend-continuation strategy on DOGE 4H, and its base backtest looked well-formed: 185 distributed trades, a 219% return, profit factor 1.52, and deployable $54.7M capacity. But the Sharpe was only 0.95 with a CI low of 0.041, and the return was bull-concentrated and fat-tail-driven (kurtosis 164), carried by a handful of pump days and two bull years. Recognizing the identical asset/timeframe/family signature as two DOGE 4H trend siblings already abandoned post-optimization this session — both of which passed in-sample and sensitivity yet died in the chronological holdout — the analyst abandoned this arm at the backtest-review gate rather than spend optimization time to re-derive the same failure.

No parameter change converts a two-bull-year, fat-tail-carried edge into a forward-robust one — when a marginal Sharpe with a CI touching zero shares the identical bull-concentrated signature as siblings that already died in the holdout, abandoning before optimization saves re-deriving the same result.

The analyst abandoned it at the backtest-review gate before optimization, judging the outcome a known dead end: no edge before selection (Sharpe 0.95, CI essentially touching zero) plus bull-concentration and fat tails meant best-of-225 optimization would inflate the Sharpe only for deflated Sharpe to correct it below the luck bar and the chronological holdout to expose the decay — exactly what happened to two DOGE 4H siblings already abandoned this session (DSR 0.0006, holdout -0.41; and DSR 0.0, holdout -1.23).

A minimal long-only, OHLCV-only trend-continuation strategy on DOGEUSDT 4-hour futures — entering long only when three filters align (close above the 50-period SMA, the SMA itself rising over the last 10 bars, and close in the upper third of its 20-bar Donchian range), then exiting via a 3x-ATR chandelier trailing stop or a 30-bar (~5-day) time stop.

The base backtest was well-populated — 185 trades over 6 years for a 219.4% total return (CAGR 20.3%), profit factor 1.52, 41% win rate, max drawdown 17.5%, and $54.7M capacity at 4.3% impact — but Sharpe was only 0.95 with a CI low of 0.041 essentially touching zero, and returns were fat-tailed (kurtosis 164, skew 10.4), carried by a few pump days and two bull years (2021 +91%, 2024 +43%) with 2020/2023/2025 flat-to-negative.