Tokyo Drift

One configuration, not three

Tokyo Drift is a structural fork of the Drift VWAP Pullback line. Drift gives Asia, London and New York a complete, independently tuned copy of every setting – three filter stacks, three sets of brackets, three breakeven engines. Tokyo Drift collapses all of it into one configuration applied to whichever sessions are enabled.

Sessions still exist and still matter, but they carry only what is genuinely theirs: the trading window, the end-of-day flatten, the direction toggles, and the trade / loss / reversal budgets. Every filter, entry rule, target, stop and trade-management setting is shared.

The reason is not tidiness. Three independently tuned copies of one strategy is three opportunities to fit the past, and a setting that looks good on one session’s few hundred trades is frequently noise. A single shared configuration has to earn its keep across every session that is switched on.

Shipped configuration

Micro Gold, 5-minute, one contract, $2.40 round turn ($1.20 per side). These are the shipped defaults – the published Pine and the numbers below are the same configuration, verified key-for-key against the parameter export.

• Entry – RSI Reset: RSI dips within a 2-bar lookback then recovers back through the midline, inside a VWAP trend requiring both the correct side of VWAP and a minimum 0.5-point slope over 10 bars.
• Context filter – RSI overbought/oversold reset, 59 / 39 with a 15-bar extreme lookback. Momentum, Efficiency Ratio, MACD Histogram Velocity, Min VWAP Distance and Max Extension are all available and currently off.
• Brackets – ATR Multiple, and deliberately asymmetric: longs take 2.5x ATR against a 3.0x stop, shorts 3.0x against 3.0x. A 5-point floor keeps a quiet ATR reading from producing a degenerate bracket.
• Anti-liquidation – entries blocked on candles wider than 15 points.
• Sessions – Asia 6:30 PM - 2:00 AM ET (3 trades / 2 losses), London 3:00 - 10:00 AM ET (3 / 1), New York 10:35 AM - 3:30 PM ET (3 / 1). One reversal per session.
• Reversal Engine – on, arming on a full stop-loss only, 3-bar wait, with its own bracket and trail: longs 35 / 45 trailing 10 points behind the peak once 20 points up, shorts 50 / 25 trailing 45 behind once 25 up.
• Breakeven and the primary trail – off. The main trade-management layer is deliberately unused in this release.

Results

Sep 2025 - Sep 2026, the full available history on this instrument:

The gap between those rows is the honest headline. Most of the year’s profit arrived in the final three months; the nine months before it were barely above break-even and carried the entire drawdown. That is a favourable regime rather than a broken strategy – the same regime shift was measured independently on the Drift VWAP line, where it survived normalising for volatility – but a figure quoted from the recent quarter alone would be selling the regime, not the strategy.

The full-year row is this version’s headline number for exactly that reason.

By session, over the full year: New York is strongest at $55.30 per trade (PF 1.92), Asia next at $29.40 (PF 1.44), London weakest at $13.64 (PF 1.19). By side, longs run $27.83 per trade at a 60.8% win rate and shorts $28.14 at 56.0% – close in expectancy despite the asymmetric targets, which is what the asymmetry is for.

The Reversal Engine contributes 70 of the 671 trades for +$2,480, at $35.43 per trade against the primary entries’ $27.17.

Verification

Entry-by-entry against a real TradingView export with the reversal engine live: 651 of 667 trades matched on timestamp and direction – 97.6% recall, 97.5% precision, with entry price exact on 649 of 651.

The residual is a known data artifact rather than an engine defect. TradingView prices the continuous gold contract slightly differently from the Databento series the engine reads, and the resulting ~0.35% ATR difference is roughly half a tick on an ATR-sized bracket – enough to flip which side of the tick the target snaps to on some trades, in both directions.

The paired parameter export and trade list are both kept in the repository, so the configuration behind these numbers is recorded rather than described.

Known gaps

• Breakeven and the primary trail are off. 38% of losing trades reached meaningful profit before reversing into the stop. Whether a breakeven captures that without cutting winners is being tested with walk-forward validation, not by selecting whichever threshold looks best in hindsight.
• London is the weak session and has not been separately addressed – by design, since fixing it in isolation is exactly the per-session tuning this fork exists to avoid.
• One instrument, one year. MES, MNQ and possibly MBT are planned next. Nothing here says how this behaves on another contract, and the brackets are ATR-scaled precisely so that question is answerable rather than assumed.