BtcSustainedMultiBarDriveContinuationLS1H

BTC Intraday Sustained Multi-Bar Drive Continuation, Long-Short (Binance USD-M, 1H, Persistence-Gated Momentum Burst, Tight Initial Stop, Trailing-Stop Winners, Discrete Capital-Capped)

A LONG-SHORT, single-instrument, pure-OHLCV intraday CONTINUATION strategy on BTCUSDT.BINANCE USD-M perpetual using 1-HOUR bars, whose DISTINCT, mechanistically-orthogonal trigger is a SUSTAINED MULTI-BAR DRIVE (persistence) rather than a SINGLE large impulse bar. It enters in the direction of a confirmed momentum burst — N consecutive same-direction 1H closes whose cumulative move exceeds a volatility threshold — to ride the continuation that follows a persistent intraday drive. This is a deliberately ORTHOGONAL complement to the promoted impulse-continuation family (the analyst's explicit directive to 'pursue genuinely orthogonal mechanisms'): the validated single-impulse trigger fires on ONE outlier bar, which is exactly the failure mode that whipsawed XRP (one-off news spikes that mean-revert), whereas requiring multi-bar PERSISTENCE filters out one-off spikes and only acts on confirmed momentum — a different trigger, a different trade population, low correlation with the single-impulse instances. It targets BTC specifically because BTC's single-impulse-1H instance is only MARGINAL (Sharpe ~0.75) — single outlier bars on BTC often chop — so a persistence-gated drive is the targeted attempt to extract a cleaner intraday-continuation signal on the deepest, most liquid, most fee-survivable major. It respects every established guardrail: validated 1H grain (NOT the fee-dead sub-1H), clean flow-driven major (NOT XRP-style news-whipsaw), single-instrument discrete enter-once/exit-once capital-capped sizing (engine-safe — NOT continuous vol-scaling/netting-flips). It is NOT a daily-trend momentum-basket variant (all banned — correlated with promoted Absolute Momentum), NOT single-asset daily trend/momentum-confluence (decayed), NOT reversion/breakout (dead), NOT funding-carry (fee-dominated), NOT options/spot (blocked), and NOT a re-expression of the single-impulse trigger. Larger cumulative drive moves clear the fee floor better than single bars, and the persistence gate yields FEWER, higher-quality trades (lower fee incidence). Core parameters kept minimal (drive length, drive magnitude, initial stop, trail).

Implements the hypothesis's orthogonal mechanism: rather than firing on ONE outlier impulse bar (the single-impulse family that whipsawed on one-off spikes), it requires multi-bar PERSISTENCE — drive_len consecutive same-direction closes whose cumulative move clears a volatility (ATR) threshold — yielding fewer, higher-quality continuation trades with a distinct trade population and low correlation to single-impulse instances. Larger cumulative drive moves clear the round-trip fee floor (~0.10%) better than single bars, and the 1.2% initial stop plus trailing winners targets per-trade edge well above fees. It respects guardrails: validated 1H grain (not fee-dead sub-1H), BTC (deepest/most fee-survivable major, where the single-impulse signal is only marginal), single-instrument discrete enter-once/exit-once capital-capped sizing (engine-safe, no continuous vol-scaling or netting flips). Routed to BINANCE USD-M futures because the strategy goes short as well as long. Leverage is left at 1.0 — risk-based sizing naturally produces under-1x notional, so leverage is honestly not referenced in sizing (avoids the leverage_set_but_unused gate).

Fundamentally broken on two independent axes. (1) NO EDGE: the multi-bar persistence continuation trigger on BTC 1H long-short is gross-losing (profit_factor 0.85, win_rate 0.39, negative expectancy, Sharpe -0.31) and negative in 5 of 6 years (only 2023 positive) — it fails in bull, bear, and chop, so varying drive_len/drive_threshold/stops in optimization cannot rescue a sub-1.0 profit factor. (2) CATASTROPHIC RISK: the account was fully LIQUIDATED (total_return -100%, max_drawdown 100%, liquidation 2026-02-13) via a fixed-stop-sizing death spiral (avg_position_pct 256% despite a max_notional_frac=1.0 cap, with daily swings of -2337%/+290%). The sizing bug amplifies but does not cause the loss — the directional signal is already negative gross of the blowup. Not worth 2 hours of optimization; the intraday momentum-continuation premise is dead on BTC 1H. (Note: the position_size/max_notional_frac cap clearly does not bind as written — if this mechanism is ever revisited, that bug must be fixed first, but it would still face a negative base edge.)

Long-short intraday continuation on BTCUSDT.BINANCE USD-M perpetual, 1H bars. The trigger is a SUSTAINED MULTI-BAR DRIVE (persistence): N consecutive same-direction 1H closes whose cumulative move exceeds an ATR-normalized volatility threshold. calculate_signal returns a continuous signed drive score (cumulative N-bar return / ATR%) every bar; should_enter gates on both the persistence count (>= drive_len consecutive same-direction closes) and the drive magnitude (>= drive_threshold ATR units), entering BUY for an up-drive and SELL for a down-drive. Exits use a tight fixed initial stop (1.2%), a trailing stop that arms after 1.5% profit and gives back 0.8% from the favorable extreme, and a strong-opposite-drive reversal exit. Sizing is discrete, risk-based and capital-capped: equity * risk_pct / (price * initial_stop), capped at 1x equity notional (no leverage).

Healthy trade count (1658 trades over 1886 days) and balanced long/short split (864/794) — the signal fires often enough that the sample is not the problem.

Code structure is sane: ATR-normalized drive score, persistence gate, trailing stop, capital-relative sizing intent.

Account was LIQUIDATED (liquidated=true, total_return -100%, max_drawdown 100%, liquidation_date 2026-02-13) — terminal blowup, not tunable underperformance.

Negative directional edge across regimes: profit_factor 0.85 (<1, gross-losing), win_rate 0.39, negative expectancy, Sharpe -0.31, Sortino -0.46.

Annual returns negative in 5 of 6 years (2020 -28%, 2021 -52%, 2022 -35%, 2024 -26%, 2025 -105%); only 2023 +12% — no regime in which the persistence-continuation trigger works.

Position-sizing death spiral: avg_position_pct 256% despite a max_notional_frac=1.0 cap — fixed-risk qty on shrinking equity pyramids into liquidation (insane daily swings: -2337%, +290%).

Fee drag is real but secondary: commission_pct_of_gross 11.69% ($69K) — the bigger issue is the gross edge is already negative.

This strategy attempted an intraday continuation trade on BTC 1H perps gated on a sustained multi-bar drive rather than a single impulse bar, pitched as an orthogonal complement to the promoted impulse-continuation family. Despite a healthy 1,658-trade sample with a balanced long/short split, it was gross-losing with a profit factor of 0.85, a 39% win rate, and negative returns in 5 of 6 years, and the account was ultimately fully liquidated on 2026-02-13 for a -100% total return and 100% max drawdown. The analyst abandoned it at the backtest-review gate on two independent axes: no directional edge in any regime, and a catastrophic fixed-stop sizing death spiral where avg position size reached 256% despite a max_notional_frac cap that failed to bind. The multi-bar persistence-continuation premise was deemed dead on BTC 1H, with the noted sizing bug requiring a fix before the mechanism could ever be revisited against its already-negative base edge.

The intraday momentum-continuation premise has no edge on BTC 1H long-short (gross-losing across bull, bear, and chop), and a fixed-risk sizing rule whose max_notional_frac cap fails to bind pyramids into liquidation on shrinking equity — both the signal and the position-sizing safeguard must be sound before any tuning is worthwhile.

It was abandoned at the pre-optimization backtest-review gate with an 'abandon' verdict, failing every criterion including liquidated=true, max_drawdown 100%, Sharpe -0.31, and profit factor 0.85; the analyst, risk, and optimization stages were never reached.

A long-short, single-instrument intraday continuation strategy on BTCUSDT.BINANCE 1H perps that entered in the direction of a sustained multi-bar drive — N consecutive same-direction closes whose ATR-normalized cumulative move cleared a volatility threshold — with a tight initial stop and trailing stop, intended as an orthogonal complement to the promoted single-impulse momentum family.

Over 1,886 days it fired 1,658 trades (864 long / 794 short) but the account was fully LIQUIDATED on 2026-02-13: total return -100%, max drawdown 100%, Sharpe -0.31, profit factor 0.85, 39% win rate, and negative expectancy (-$65/trade), with annual returns negative in 5 of 6 years (only 2023 was positive at +12%).