Pipsgrowth EX18004 TrendFollow

Pipsgrowth EX18004 TrendFollow hunts directional moves on XAUUSD M5 by combining the ADX/DI cross on the last closed bar with a 7-state market regime classifier that filters out environments where the cross is most likely to fail. The EA is built to be patient. A new setup only triggers on bar close, and only after the regime layer confirms the market is tradable — so on a quiet Asian session with ADX under 15 and a flat ATR percentile, the bot simply logs the state and waits.

The signal engine is the standard ADX(14) module with three buffers: ADX main, +DI, and −DI. The InpADXThreshold input (default 25) is the floor; if ADX is below it, the cross is ignored regardless of how clean the DI line looks. With InpRequireDICross = true, the EA only fires when the DI lines actually cross on bar close, not on a persistent +DI>−DI gap. A long fires when +DI crosses above −DI and ADX≥25, with a signal confidence of 70. A short fires on the mirror image. If the cross requirement is disabled, any +DI>−DI majority with ADX above threshold logs a 55-confidence bias instead.

Before a position opens, ClassifyRegime() reads ADX, ATR(14) percentile over a 100-bar lookback, and Bollinger Band(20, 2.0) width, then assigns one of seven states: REG_STRONG_TREND, REG_WEAK_TREND, REG_RANGE, REG_COMPRESS, REG_BREAKOUT, REG_CHOPPY, or REG_NONE. Two of those states are hard-blocked: REG_CHOPPY (ADX under 15) and REG_COMPRESS (ATR below the 40th percentile and BB width under 50 points). The remaining four pass to the next stage, each with a confidence output — STRONG_TREND at 80, BREAKOUT at 85, WEAK_TREND at 55, RANGE at 40. The CHOPPY and COMPRESS cases are logged with the regime code and confidence score, but no order is submitted.

ConfirmEntry() then runs three additional checks. The HTF ADX (default PERIOD_H1) must agree: if the H1 ADX is above InpADXThreshold-5 and the H1 DI lines oppose the trade direction, the entry is rejected with the reason HTF против. A 14-period RSI midpoint filter rejects longs below 50 and shorts above 50. Finally, spread must be under InpMaxSpreadPts (300 points = 30 pips for XAUUSD) and the implied R:R (TP-mult / SL-mult = 4.0/2.0 = 2.0) must clear 1.0.

The SL is computed from ATR(14) × InpATRSLmult (default 2.0×), then run through EnforceStopsLevel() to push the distance above the broker's minimum stop distance plus a 5-point buffer. The TP is the SL distance × (InpATRTPmult / InpATRSLmult) = 2× SL by default, so the design ratio is 1:2 R:R. Lot sizing is risk-based: CalcLot() divides g_effectiveCapital × InpRiskPercent% (default 0.5% of equity, or capped at InpCapAmount if set) by the per-lot risk at the calculated SL. If risk is 0, it falls back to InpFixedLot (0.01). Volumes pass through NormalizeVolume() which floors to the lot step and clamps to the symbol's min/max.

Position management runs every tick through ManageExits(). The first check is an opposite-signal exit: if the regime and DI cross produce a long signal while only sells are open, the sell side is closed via CloseSide(POSITION_TYPE_SELL, "opp-signal"). The second is a regime-change exit: if ClassifyRegime() returns CHOPPY on any tick, all open positions for this magic are closed with the reason choppy. The third is a time stop: positions older than MAX_BARS_IN_TRADE (200 bars = roughly 16.7 hours on M5) are force-closed with the time-stop reason.

The trailing and break-even logic both run on the same per-position loop. When unrealized profit reaches 1R (defined as ATR × InpATRSLmult), the stop is ratcheted to open + 0.05×ATR for longs or open − 0.05×ATR for shorts — a tight break-even that removes the loss case but still gives the trade room to breathe. The trail is set to ATR × 2.5 from the current bid/ask, and the EA only ever moves the stop in the favorable direction (MathMax for longs, MathMin for shorts, with a 0-SL safety branch). Each modify call is wrapped in TryModify_EX18004() which retries up to 3 times on requote/timeout with 100ms sleeps.

A one-shot partial close fires when profit reaches 0.5 × the TP distance. If the resulting half volume clears the broker's minimum lot and 1.5 × min_lots, half the position is closed via TryClosePartial_EX18004() (3 retries, 200ms sleeps on transient) and the ticket is pushed into g_partialClosedTickets[] so it never fires again on the same trade.

The pyramid layer is off by default. With InpPyramidMaxLevels > 0 (max 5), ManageScaling() runs after ManageExits(), checks that all existing positions in the direction are profitable by 0.5 × ATR, then re-uses TryEntry() to add a layer. The scaling inherits all the same regime, signal, confirm, and no-trade gates as the original entry, so a pyramid add on a weak regime is impossible.

Safety sits across the whole stack. NoTrade() blocks entries if the kill switch is on, if spread exceeds 300 points, if the server hour is outside 7-21, on weekends, if the effective capital drops below InpCapFloor (default $50), if realized daily loss has hit 3% of effective capital, if the weekly loss has hit 9%, if InpMaxConcurrent (default 1) positions are already open, or if g_consecLosses ≥ 3 and we are still inside the 5-bar cooldown window. All PnL tracking walks HistoryDealsTotal() for deals matching this magic and this symbol only — no other EAs are touched. The dry-run switch (InpDryRun = true by default) lets you log every signal that would have fired without sending an order, which is how you should first run this EA on a live terminal.

The OnTester formula is (STAT_PROFIT × STAT_PROFIT_FACTOR) / (1 + STAT_BALANCE_DDREL_PERCENT), which weights the optimization criterion toward both net return and the profit factor while penalizing the relative balance drawdown. If the test has fewer than 30 trades, OnTester returns 0 to suppress noisy optimizations.

What to expect in backtest: because the EA only fires on a closed-bar ADX/DI cross and the regime gate rejects CHOPPY and COMPRESS, the trade frequency on XAUUSD M5 is moderate — typically a handful of setups per London-NY session, fewer during the Asia overlap. The default 1:2 R:R and 0.5% risk per trade means a long quiet trend produces a steady string of small wins plus the occasional 2R home run, while a choppy week will mostly produce cooldown-driven silence rather than losses. The partial close at 0.5×TP locks in roughly 1R on every winner regardless of where the final exit lands, which is what gives the equity curve its smoother look in optimizer runs.