DogeThirtyMinKeltnerRangeTradingLongShort

DOGE 30-Minute Keltner Channel Statistical Range Trading on DOGEUSDT.BINANCE (Long-Short, Market-Making Adjacent)

A long-short, short-horizon, market-making-adjacent statistical range-trading strategy on DOGEUSDT.BINANCE — confirmed-testable per recent pipeline additions (DOGEVolContractionSqueezeLong, DOGEVolatilityTargetedLong, DOGEADACointegrationPairsReversion60Day all using DOGE bars successfully). Responds DIRECTLY to the analyst's repeatedly-flagged unfilled mechanism class: 'Market making / spread capture (no significant exploration yet)'. Structurally distinct from every current portfolio strategy — those are multi-timeframe directional or multi-day mean-reversion; this is short-horizon (30-min entries, 2-8 hour holds) symmetric range trading. Mechanism (Keltner 1960, Chester Keltner; Linda Raschke 1996 systematic adaptations; AmberData crypto microstructure research 2023): DOGE's higher idiosyncratic volatility relative to its trend rate creates a documented persistent range-trading regime intraday. Keltner Channels (EMA ± multiplier × ATR) define statistically-anchored upper/lower bounds. Liquidity providers earn the bid-ask-equivalent spread by buying at the lower channel and selling at the upper channel — this strategy systematizes that approach. The 30-min timeframe is fast enough to capture intraday reversion (multiple touches per day in active periods) but slow enough to avoid the sub-minute microstructure noise that killed prior intraday pattern strategies. NO trend filter — this is symmetric range trading by design (the trend filter on a market-making strategy defeats its purpose by halving setup frequency). Uses ONLY DOGEUSDT.BINANCE 30-min bars + ATR — the simplest possible OHLCV dependency on a confirmed-working instrument. Fills critical gaps: (1) NEW mechanism class (market-making/range-trading, not present in portfolio), (2) SHORT-HORIZON timeframe (mid_30m_2h at 10.5% — under-quota), (3) LONG-SHORT direction (15.1% → toward 45%), (4) confirmed-testable instrument (avoids the audit_stale plague across majors), (5) single-instrument simplicity (minimizes auto-collection dependency surface).

Directly fills the analyst's repeatedly-flagged unexplored mechanism class — market making / spread capture — as short-horizon symmetric range trading, structurally distinct from the portfolio's multi-timeframe directional and multi-day mean-reversion strategies. The Keltner channel (EMA ± ATR) gives statistically-anchored bounds; buying the lower band and selling the upper band systematizes the liquidity-provider spread, and the deliberate absence of a trend filter keeps it symmetric (a filter would halve setups and defeat the purpose). The signal is the close-to-EMA distance in ATR units, computed continuously every bar (clamped ±20, never frozen) and thresholded at ±atr_mult in should_enter for both BUY and SELL — verified through the actual Layer-2 proxy (all six scenarios non-frozen, 460 unique, no exceptions) and against realistic range-bound+spike data where the z reached both bands and fired both directions (5 BUY, 1 SELL). The 30-minute timeframe captures intraday reversion while avoiding sub-minute microstructure noise; DOGEUSDT.BINANCE is a confirmed-testable instrument. It fills four portfolio gaps (new MM mechanism class, short-horizon timeframe, long-short direction, confirmed instrument) with a single-instrument OHLCV-only footprint. Leverage is 1.0 (the USD-M MARGIN account permits the short leg at 1x), so sizing references no margin leverage and avoids the inert-leverage gate; risk-capped sizing off the 3% stop controls per-trade risk.

Decisively money-losing with no edge and a fundamentally flawed premise. Over an ample 2,946-trade / 5.8-year sample: Sharpe -1.44 with the ENTIRE confidence interval below zero (-2.09 to -0.78), probabilistic_sharpe 0.0, profit_factor 0.871, total_return -86.4%, max_drawdown 88.4% (CI to 97%), negative in 6 of 7 years and accelerating to ruin (2024 -38.5%, 2025 -43.1%, 2026 -37.3%). The Keltner band-fade has an inverted payoff -- 59.9% win rate but avg_loss 549 is ~1.7x avg_win 320 (skew -2.37, tail_ratio 0.57) -- because DOGE's explosive meme-driven trends run over a range-fade: band touches are breakouts as often as reversions, so the strategy wins small reverting to the midline and loses big when price keeps going. Crucially, the 'market-making / spread-capture' thesis is mechanically wrong: real MM EARNS the spread by POSTING limit orders, but this strategy crosses the spread with MARKET orders at the bands, so it PAYS fees + impact ($51.6k commission + $49.9k impact, 9.13% of gross) -- it is a fee-bleeding fade, not spread capture. No parameter optimization fixes a 0.871-PF, -86%, 88%-DD, wholly-negative-Sharpe-CI fade; DOGE does not range-revert at 30-min, it trends, and the fees compound the bleed. Do not spend optimization compute. FAILURE PATTERN: symmetric Keltner/Bollinger band-fade 'range trading' (mislabeled as market-making) on a trending meme coin (DOGE 30-min) is a money-losing inverted-payoff strategy -- high win rate, fat left tail, sub-1 PF, wholly-negative Sharpe CI -- because the asset trends through the bands and the market-order execution pays rather than earns the spread; genuine market-making requires posting limit orders (maker fills), which the FactoryStrategy market-order model cannot represent, so spread-capture strategies are not viable in this harness as constructed.

Long-short, market-making-adjacent symmetric range-trading strategy on DOGEUSDT.BINANCE 30-minute bars (pure OHLCV, single instrument). It builds a Keltner channel (EMA(20) ± 2×ATR(20)) and fades band touches: BUY when the close is at/below the lower band, SELL when at/above the upper band — no trend filter, by design. Each position exits on mean-reversion back toward the EMA midline (within 0.3 ATR), a 3% adverse stop, or a 16-bar (~8h) time stop. Capital-relative, risk-capped sizing. Leverage 1.0.

Clean implementation, correct long/short routing (1450 long / 1496 short), ample sample (2946 trades), metrics_reliable=true; novel mechanism-class intent (range trading / MM-adjacent)

Decisively money-losing: Sharpe -1.44 with ENTIRE CI negative (-2.09 to -0.78), probabilistic_sharpe 0.0, profit_factor 0.871, total_return -86.4%, max_drawdown 88.4%

Negative in 6 of 7 years, accelerating to ruin (2024 -38.5%, 2025 -43.1%, 2026 -37.3%)

Inverted payoff: 59.9% win rate but avg_loss 1.7x avg_win (skew -2.37) -> the band-fade gets run over by DOGE's explosive meme trends

Flawed premise: it crosses the spread with market orders (PAYS fees + impact: $51.6k commission + $49.9k impact, 9.13% of gross), so it cannot 'capture the spread' -- it's a fee-bleeding fade, not market-making

DogeThirtyMinKeltnerRangeTradingLongShort set out to fill the unexplored market-making/spread-capture mechanism class with a symmetric Keltner band-fade on DOGE 30-min, buying the lower band and selling the upper to systematize liquidity provision. The implementation was clean and the sample ample, but the result was ruinous: -86.4% with a -1.44 Sharpe whose entire CI sat below zero, an 88.4% drawdown, and acceleration toward ruin in recent years. Two flaws sealed it: DOGE's explosive trends run over a range-fade (a 60% win rate undone by losses 1.7x the wins), and the 'spread-capture' premise is mechanically false — crossing the spread with market orders pays ~9% of gross in fees and impact rather than earning the spread. The reviewer abandoned it pre-optimization, drawing the meta-conclusion that genuine market-making requires limit-order posting the harness cannot model, so spread-capture strategies are not viable here as constructed.

A symmetric Keltner/Bollinger band-fade mislabeled as 'market-making' on a trending meme coin is a money-losing inverted-payoff strategy (high win rate, fat left tail, sub-1 PF, wholly-negative Sharpe CI) because the asset trends through the bands and market-order execution pays rather than earns the spread; genuine market-making requires posting limit orders for maker fills, which the FactoryStrategy market-order model cannot represent, so spread-capture strategies are not viable in this harness as constructed.

The analyst abandoned it at the backtest-review gate for a fundamentally flawed premise plus no edge: the 'spread-capture' thesis is mechanically wrong because the strategy crosses the spread with market orders (paying fees rather than earning the spread like a limit-posting market-maker), and DOGE trends through the bands rather than range-reverting at 30-min — so the band-fade has an inverted payoff (small reverting wins, fat left-tail losses) that no parameter tuning can fix.

A long/short, single-instrument, OHLCV-only 'market-making-adjacent' symmetric range-trading strategy on DOGEUSDT.BINANCE 30-minute bars using Keltner channels (EMA ± 2×ATR): buy at/below the lower band and sell at/above the upper band to fade moves and capture the spread, with no trend filter (symmetric by design), exiting on reversion to the midline, a stop, or a ~2-8 hour time stop — intended to fill the unexplored market-making/spread-capture mechanism class.

It was decisively money-losing toward ruin over an ample 2,946-trade / 5.8-year sample: -86.4% total return (CAGR -26.7%), Sharpe -1.44 with the entire CI below zero (-2.09 to -0.78), probabilistic_sharpe 0.0, profit factor 0.871, 88.4% max drawdown (CI to 97%), negative in 6 of 7 years and accelerating (2024 -38.5%, 2025 -43.1%, 2026 -37.3%) — with a 59.9% win rate but avg_loss ~1.7x avg_win (skew -2.37) and fees+impact at 9.13% of gross ($51.6k + $49.9k).