BtcLiquiditySweepMomentumContinuationLS

BTC Perp 15-Minute Liquidity-Sweep Momentum Continuation (Stop-Run Breakout, Long-Short, Asymmetric Trailing)

A short-horizon, long-SHORT, single-instrument microstructure MOMENTUM-CONTINUATION strategy on BTCUSDT.BINANCE perpetual futures using 15-MINUTE bars and OHLCV-ONLY data. This is the ANALYST-SANCTIONED mirror of the just-abandoned liquidity-sweep FADE, whose fully-sampled −99.70% result (8,837 trades, negative every year, Sharpe −2.49) empirically PROVED that BTC 15m sweeps of swing highs/lows CONTINUE far more often than they reverse. The analyst's explicit recommendation was to propose the inverse fresh — trade WITH the breakout continuation, gated on the per-trade edge surviving 15m fee drag — and confirmed the BTC 15m OHLCV path is excellent for trade count with no event-sparsity and no fragile supplementary-data dependency. Mechanism: when price decisively breaks a recent swing high (or low), it takes out the stop-loss / breakout orders clustered just beyond that level, triggering a stop cascade plus breakout-chaser momentum that tends to EXTEND over the next several bars. The strategy enters in the direction of the break. Two deliberate design choices fix the exact reasons the fade died and differentiate this from a naive 15m breakout: (1) SELECTIVITY — it only acts on decisive closes BEYOND the level by an ATR-scaled margin (significant stop-runs, not the marginal same-bar wick pokes the fade traded), so the per-trade move is large relative to the ~0.10% round-trip fee; (2) ASYMMETRIC PAYOFF — a tight stop at the reclaimed level versus a wide ATR trailing stop that rides the continuation, producing avg_win >> avg_loss (the opposite of the parent fade's near-symmetric $57-win/$62-loss profile that guaranteed negative expectancy). Simple (4 core parameters), single-asset, deep multi-year history; fills the under-represented short-horizon (2.2%) and long_short buckets. It is mechanistically distinct from this session's pending 1H vol-expansion-momentum seed via its specific swing-level liquidity/stop-run trigger and 15m horizon.

Implements the analyst-sanctioned INVERSE of the falsified liquidity-sweep fade: because BTC 15m sweeps of swing highs/lows were empirically shown to CONTINUE rather than reverse, this trades WITH the break. The two fixes that killed the fade are coded explicitly. (1) SELECTIVITY: should_enter only fires on a close BEYOND the swing level by breakout_atr_margin x ATR — significant stop-runs, not marginal same-bar wick pokes — so the per-trade move is large relative to the ~0.10% round-trip 15m fee, addressing the fee-drag that sank the 8,837-trade fade. (2) ASYMMETRIC PAYOFF: should_exit uses a tight stop at the reclaimed breakout level (initial risk ~= the ATR margin) combined via max()/min() with a wide 3x-ATR trailing stop that ratchets with the running high/low, so avg_win >> avg_loss — the opposite of the parent's near-symmetric $57/$62 profile. calculate_signal returns a CONTINUOUS (close - swing-mid)/ATR channel score that varies every bar (defeats the Layer-2 frozen-signal check) and carries breakout conviction, while the discrete long/short entry lives in should_enter. Robust entry-state tracking reads the actual fill (avg_px_open) and the pending breakout level/direction, and maintains trailing extremes regardless of fill timing. Sizing is capital-relative and risk-anchored to the tight stop with a min-stop floor so tiny margins can't oversize, plus a notional cap. OHLCV-only, single instrument, long+short on Binance USD-M futures (shorts required -> futures venue, lower 0.05% taker than spot); leverage=1.0 and sizing never multiplies by leverage, so it is not flagged leverage_set_but_unused. 4 core parameters as the hypothesis specifies, deep multi-year BTC 15m history, no supplementary-data dependency.

Fundamentally broken — the strategy LIQUIDATED the account (total_return -100%, max_drawdown 100%, liquidation_date 2021-01-28, monthly 2021-02 -9028% with impossible -284%/-201% single days from 143% avg position). This is the 'inverse' of the already-abandoned liquidity-sweep FADE, and it confirms the decisive lesson: BOTH directions of the BTC 15m swing-level break lose, so the move is noise net of the ~0.10% round-trip fee and has no edge either way. The design DID deliver the intended asymmetric payoff (avg_win $255 ~= 1.9x avg_loss $134), but the breakout-continuation win rate is only 26.5%, giving EV ~ -$31/trade (profit_factor 0.69) BEFORE leverage; commissions then take 25.5% of gross (+$57k impact). The hypothesis's premise is refuted: the parent fade's -99.7% did NOT prove sweeps 'continue' — the fade lost to symmetric payoff + fees, and the continuation loses because the directional hit rate is too low to monetize even a 1.9x payoff. Over-leveraged tight-stop sizing then converts the negative EV into a full wipeout. No parameter sweep can rescue a strategy that liquidates, is negative-EV pre-leverage, and is fee-dominated, and the optimizer cannot run because the account is dead for ~80% of the window. META-NOTE FOR RESEARCH LEAD: stop proposing the BTC 15m swing-break liquidity-sweep family in EITHER direction (fade or continuation) — both have now been fully tested to -100%/liquidation, establishing that a 15m swing-level break on BTC carries no edge that survives the fee drag. FAILURE PATTERN: leveraged long/short 15m liquidity-sweep momentum-continuation on BTC liquidates the account — selectivity + asymmetric payoff cannot overcome a 26.5% continuation hit rate, 25%+ fee drag, and tight-stop over-leverage; this is the mirror confirmation that intraday swing-break strategies on majors are sub-fee noise.

Long-short microstructure momentum-CONTINUATION on BTCUSDT.BINANCE USD-M perpetual, 15-minute OHLCV-only bars. Each bar it computes a 20-bar prior swing high/low and ATR(14). It enters WITH a decisive stop-run breakout: BUY when close exceeds the swing high by breakout_atr_margin x ATR (0.5 ATR), SELL when close breaks the swing low by the same ATR-scaled margin. Payoff is deliberately asymmetric: a TIGHT initial stop at the reclaimed breakout level (swing high/low) versus a WIDE ATR trailing stop (3x ATR off the running extreme) that rides the continuation, plus a 24-bar (~6h) time stop. Position sizing is risk-based to the tight stop: equity x 1% / (breakout_atr_margin x ATR), capped at 95% of equity notional. leverage=1.

Achieved the intended asymmetric payoff (avg_win $255 ~= 1.9x avg_loss $134), unlike the near-symmetric parent fade.

Ample signals / no event-sparsity (5155 entries) and clean OHLCV-only execution.

ACCOUNT LIQUIDATED 2021-01-28 (total_return -100%, max_drawdown 100%) with impossible -284% day / -9028% month from 143% avg position — ruinous over-leverage from tight-stop risk sizing.

Negative edge even with the favorable payoff: win_rate only 26.5%, so EV ~ -$31/trade; profit_factor 0.69.

Fee-dominated at 15m: commission 25.5% of gross plus $57k impact erases any residual edge.

Refutes its own premise: trading WITH the break loses just like the fade — the BTC 15m swing-break is noise net of fees, with no edge in either direction.

Post-liquidation flat for ~80% of the window, so optimization cannot run meaningfully.

BtcLiquiditySweepMomentumContinuationLS was the analyst-sanctioned mirror of an abandoned sweep-fade, trading WITH BTC 15m breakouts on the premise that the fade's -99.7% proved sweeps continue, with selectivity and an asymmetric tight-stop/wide-trail payoff to fix the fade's flaws. It did achieve avg_win ≈ 1.9x avg_loss, but the continuation hit only 26.5% of the time — negative EV before leverage — and fees took a quarter of gross, while 143% over-leverage drove the account to a -100% liquidation in early 2021. The analyst concluded the premise was refuted: both fade and continuation lose, so the 15m swing-break is sub-fee noise, with a meta-note to stop proposing the family in either direction. It ended after one iteration as abandoned, never advancing to optimization or risk review.

Both directions of the BTC 15m swing-level break now lose to -100%/liquidation — the fade lost to symmetric payoff plus fees, and the continuation loses because a 26.5% hit rate cannot monetize even a 1.9x asymmetric payoff against ~25% fee drag — establishing that a 15m swing-break on BTC is sub-fee noise with no edge in either direction, and that selectivity and favorable payoff cannot rescue a low directional hit rate or tight-stop over-leverage.

It was abandoned at the pre-optimization backtest-review gate (verdict: abandon) as fundamentally broken — the breakout-continuation hit rate (26.5%) was far too low to monetize even a 1.9x payoff, leaving negative EV before leverage that fees compounded and tight-stop over-leverage converted into a full wipeout — so optimization (impossible with the account dead ~80% of the window) and all later stages were never reached.

A long/short single-instrument microstructure momentum-continuation on BTCUSDT.BINANCE perps (15-minute bars, OHLCV-only) that traded WITH liquidity-sweep breakouts — entering long when price closed beyond a 20-bar swing high by an ATR-scaled margin (and short on the mirror), with a tight stop at the reclaimed level and a wide ATR trailing stop for asymmetric payoff — the analyst-sanctioned inverse of a just-abandoned sweep-fade, designed for selectivity and avg_win >> avg_loss.

It liquidated the account on 2021-01-28: total return -100%, max drawdown 100% (impossible -284% day / -9028% month from a 143% average position), and even before leverage the edge was negative — a 26.5% win rate gave expectancy -$31/trade and profit factor 0.69 despite achieving the intended asymmetric payoff (avg win $255 ≈ 1.9x avg loss $134), with commissions consuming 25.5% of gross plus $57k impact.