Pipsgrowth EX18067 TrendFollow

The Momentum Stacker runs on the M5 chart but builds its decision from a vertical stack of three timeframes. Three independent Hull-MA pairs are pulled in via iMA(... MODE_LWMA ...) — half-period LWMA and full-period LWMA — for the M1 (20-bar), M5 (30-bar), and M15 (50-bar) charts, and the HullDir() helper compares 2×LWMA(period/2) − LWMA(period) across two consecutive closed bars to return +1 (rising), −1 (falling), or 0 (flat). Those three directional votes feed a six-slot consensus that also scores MACD histogram acceleration on the M1 chart (the M1 histogram value at bar 0 must exceed the value at bar 1, and the sign must match the trade direction), an M1 ATR expansion bar (the 14-period M1 ATR at bar 0 strictly exceeds the value at bar 1), and a tick-volume surge on the M1 chart where the latest 1-minute bar is at least 1.3× the rolling 20-bar average. The signal fires only when at least 4 of these 6 votes (the InpMinConfidence threshold) agree, and the check uses shift=1 so it never re-paints on the still-open bar.

The regime layer sits ahead of the signal and uses the same 14-period ADX and 14-period ATR that later drive the SL distance, plus a 20-period / 2.0-dev Bollinger Band whose width is measured as (upper − lower) / close. ClassifyRegime() first tests for a quiet, tight market (Bollinger width ≤ 0.0150 of price AND ATR percentile ≤ 25 → REG_COMPRESS). Then it tests for REG_STRONG_TREND (ADX ≥ 25 AND ATR percentile ≥ 75). If ADX is at least 18 with ATR percentile ≥ 50, the regime is REG_WEAK_TREND. A high ATR percentile alone with no trend filter yields REG_EXPAND, and a quiet ADX below 15 returns REG_CHOPPY. Three of those six states — REG_RANGE, REG_CHOPPY, REG_COMPRESS — are blockers: the entry gate at the top of OnTick() returns early when g_curRegime lands in any of them. Strong trend, weak trend, and expand are tradeable; momentum is allowed to lead price in any of them.

A set of confirm filters and a set of no-trade gates run before every entry. The confirm layer requires ConfirmHtf() to pass (the M15 close must be on the correct side of the M15 EMA(50)), ConfirmCandleClose() to pass (the most recent M5 candle must be a bullish close for a long, bearish for a short), and ConfirmMinRR() to pass (the projected TP distance versus the SL distance must clear InpMinRR=1.3; when TP is zero, the check uses 2×SL as a stand-in to keep the math well-defined). The no-trade layer checks effective capital against the floor, realized PnL against the daily and weekly loss caps, a 15-minute post-loss cooldown, the current spread against DEF_MAX_SPREAD_PIPS=2.5, the server hour against the 7–20 session window, weekend flags, and the broker's SYMBOL_TRADE_MODE flag.

Risk and sizing are bound together. The risk budget is 0.5% of the effective capital per trade, hard-capped at InpMaxLot=1.0. The effective capital is min(account equity, InpCapitalCapAmount) when the cap input is non-zero, otherwise pure account equity, and a InpCapitalCapFloor=50.0 guard disables entries entirely when effective capital falls below $50. The CalcLots() function converts the dollar risk into lots by dividing riskAmt by the per-lot stop-loss cost, which is derived from the SL distance times g_tickVal / g_tickSize (the broker's tick-value-to-tick-size ratio). Free margin is verified with OrderCalcMargin() before SendOrder() is invoked.

Position management runs in four phases. The first is break-even: at 1.0R profit the stop is ratcheted to open + 1 pip for longs and open − 1 pip for shorts, but only if the move actually tightens the stop (one-shot). The second is partial take-profit: at 1.5R, half the position is closed via TryClosePartial_EX18067 and the rest stays in play. The third is the trailing stop, computed as g_curAtr × InpATRTrailMult = 2.5 × ATR(14) behind the current bid or ask, and it only engages once the position is at least 0.5R into profit; the trail is forward-only and never moves backward. The fourth is the exit cluster: opposite-signal close (the latest EvaluateSignal() direction is against the position), regime-change close (the regime flipped to range, choppy, or compress), and a MAX_BARS_IN_TRADE=240 time stop (M5 × 240 ≈ 20 hours).

Pyramiding is wired but disabled by default. DEF_PYRAMID_ENABLED=false means a long signal is rejected outright when there is already a long position on the book. If the operator flips it on, the same PyramidMaxLevels=3 ceiling applies and a new entry requires the current price to be at least 1.5 × ATR(14) away from the most recent leg in the same direction. The directional exposure caps — DEF_MAX_SYMBOL_LONG=3 and DEF_MAX_SYMBOL_SHORT=3 — apply independently of pyramid mode and sit on top of the magic-wide InpMaxConcurrent=5 limit. A DuplicateEntryGuard (using g_lastEntryBarTime) also enforces MIN_BARS_BETWEEN=1 between consecutive entries on the same magic.

The OnTester() custom fitness score is (netProfit × profitFactor) / (1 + balanceDDPercent), returning 0 when the trade count is below 30 or drawdown is non-positive. The order comment is hardcoded to Psgrowth.com Expert_18067, deviation to 30 points, and the filling mode is set per-symbol via SetTypeFillingBySymbol. All three order-modify helpers (TryClose_EX18067, TryClosePartial_EX18067, TryModify_EX18067) retry three times on requote, timeout, price-off, and price-changed retcodes, with 200ms sleeps on close attempts and 100ms on modifications. As shipped, InpDryRun=true means the EA logs would-be entries to the journal only; you must flip it to false to send live orders, and the global InpKillSwitch is an extra hard stop that returns immediately from OnTick().