BTC Intraday VWAP-Anchored Continuation Trend (Single-Instrument USD-M, Long-Short, 1H, Ride Excursions Away From Daily VWAP, Cut on VWAP Re-Cross, Trailing-Stop Winners)

A long-SHORT, SINGLE-INSTRUMENT, pure-OHLCV intraday TREND-CONTINUATION strategy on BTCUSDT.BINANCE USD-M perpetual futures using 1-HOUR bars. It is the MECHANICAL INVERSE of the factory's just-falsified VWAP-excursion mean-reversion fade, derived directly from that backtest's measured result rather than a new hypothesis. The fade (fade the move away from the daily-anchored VWAP, betting on reversion) was anti-predictive on BTC: total_return -90.5%, win rate 52% but avg_loss ~2x avg_win, losing every year 2020-2026 -- BECAUSE, as the analyst diagnosed, 'crypto intraday excursions TREND through the VWAP anchor rather than revert.' That is a directional signal pointing the OPPOSITE way: riding the excursion as CONTINUATION (go LONG when price breaks decisively above the daily VWAP, SHORT when below) inverts the payoff to avg_win ~2x avg_loss, and the per-trade math nets positive after fees (the fade's -$40 expectancy was gross-negative PLUS fee drag; the inverse is gross-positive MINUS the same drag). EVERY design choice respects this session's kills: (1) DIRECTION = continuation/trend, NOT a fade -- mean-reversion fades are now closed/anti-predictive on crypto intraday across the entire vol curve (VWAP-MR SOL -80%, BTC -90.5%); the surviving direction is trend. (2) PURE OHLCV, SINGLE INSTRUMENT -- immune to the SIX-times-killed supplementary-data wall (OI/taker/L-S/liquidations) and the basket data-collection wall (thin alts abort runs). (3) BTC -- the asset with the strongest MEASURED inverse signal (-90.5% fade => strongest continuation), lowest intraday noise, deepest data. (4) 'CUT LOSSES, LET WINNERS RUN' structure -- exit FAST when price re-crosses the VWAP (the intraday trend is over) and TRAIL the winners, which manufactures the avg_win > avg_loss asymmetry the fade lacked and that lets intraday continuation clear fees. (5) NOT a generic Donchian/EMA breakout (which died on daily/4H majors in the chop regime) -- this is an intraday VWAP-anchored continuation with direct empirical support from this factory's own data; intraday and daily are different regimes. (6) NOT a funding-confirmed-trend clone (ETH/BTC-only, and this uses no funding), NOT cross-asset ratio, NOT supplementary-primary, NOT engine-incompatible continuous-spot-rebalance. Intentionally minimal: 3 tunables (entry_z, trail_atr_mult, vol_lookback).

Implements the mechanical inverse of the factory's falsified VWAP-excursion mean-reversion fade, derived from that backtest's measured -90.5% on BTC (crypto intraday excursions trend through the daily VWAP anchor rather than revert), using the standard FactoryStrategy hooks. It reuses the validated daily-anchored-VWAP machinery (UTC-midnight reset via integer ns math, mean-of-closes fallback for zero-volume bars, rolling std of close-minus-VWAP for the z-score) but inverts the entry direction to continuation: z >= +entry_z -> LONG, z <= -entry_z -> SHORT. The returned z-score is a continuous per-bar signal so Layer-2 frozen-signal detection passes, while the threshold entry (with a MIN_DAY_BARS=3 warmup and a 22:00 entry cutoff) and the exits live in the gate methods. The exit structure is what manufactures the avg_win > avg_loss asymmetry the fade lacked: a FAST VWAP re-cross cut (a long exits the instant price falls back to/through VWAP, a short when it rises back through) cuts losers near the anchor, while a 2x-ATR trailing stop from the position's running high/low lets a runner continue and only exits on a pullback from its extreme; a 23:00 UTC forced flat ensures no position is carried across the midnight VWAP-anchor reset (stale-anchor risk). Sizing is capital-relative (30% of equity) clearing the BTCUSDT $100 min-notional; the ~0.10% round-trip futures fee is cleared by riding multi-z continuation moves with trailed exits. Long+short on BTC (deepest data, strongest measured inverse signal); pure OHLCV single-instrument avoids the supplementary-data and multi-leg-basket walls. leverage stays 1.0, referenced only as a non-amplifying sizing multiplier so the unused-leverage gate does not apply. Three tunables keep the optimization surface minimal.

The 'mechanical inverse of the falsified VWAP-MR fade' premise is empirically falsified: the inverse is ALSO a loser. profit_factor 0.84 (<1.0), Sharpe -0.84, expectancy -$17/trade, losing every year (2022 -48.9%, 2023 -66.2%, 2024 -147.3%), and a full LIQUIDATION on 2024-10-08 (total_return -100%, max_drawdown 100%). This is the inverse-of-a-loser fallacy: the prior fade's real cost was not merely 'wrong direction' but FEES, and fees are paid in BOTH directions, so flipping the sign produces another fee-bleeder (commission_pct_of_gross 22.11%, turnover 307x, $144k commission at 1H) rather than a mirror winner. Compounding this: (a) ~4x over-exposure -- avg_position_pct 118.4% vs the intended position_pct 0.30 -- because the discrete entry re-fires while a position is open at 1H, stacking size and driving the liquidation; and (b) physically-impossible late-2024 daily prints (-915% on 2024-11-01, -227% on 2024-10-27, +187% on 2024-10-29) with return_kurtosis 451 and a -9393% month, confirming the over-leverage. Not optimize: a return series dominated by impossible bars and ending in liquidation cannot be optimized. Not iterate: fixing the entry-stacking bug would only reveal a PF-0.84 signal that still bleeds 22% of gross to fees -- there is no positive-expectancy region to tune toward, and the three tunables (entry_z/trail_atr_mult/vol_lookback) cannot fix a sub-fee edge. FAILURE PATTERN: inverting the DIRECTION of a falsified intraday strategy does NOT manufacture an edge when the original's loss was driven by trading costs -- both the fade (PF<1, -90.5%) and its continuation inverse (PF 0.84, liquidated) lose because the per-trade edge is below the ~0.10% round-trip x 307x-turnover fee drag (22% of gross). Plus the discrete-entry over-exposure/stacking bug (avg_position 118% vs a 30% target, impossible prints, liquidation) recurs on this mechanism exactly as on the BNB/XRP single-asset TSM siblings. An 'empirically derived inverse' is not a new edge; it is the same fee-dominated non-edge with the sign flipped.