Pipsgrowth EX18024 TrendFollow

Pipsgrowth EX18024 is a trend-following EA built around an adaptive HMA + RSI + Heikin-Ashi signal stack, tuned specifically for short-term XAUUSD action on a 5-minute chart. The skeleton is the same 12-layer architecture used across the EX18 family (REGIME, SIGNAL, ENTRY, CONFIRM, NO-TRADE, CAPITAL CAP, RISK, SIZING, MANAGE, EXIT, SCALING, OnTester), but every layer is wired for gold volatility rather than FX majors. A lot of the inputs are hidden behind the seven #define constants at the top of the source (HA_LOOKBACK, MAX_PYRAMID_LEVELS, PYRAMID_ATR_SPACING, MAX_DAILY_TRADES, MAX_BARS_IN_TRADE, SESSION_START_HOUR/END_HOUR) which gives the EA a coherent risk profile without forcing the user to think about them.

The signal layer is the unusual part. Rather than a plain MA crossover, the EA reconstructs a Hull MA(34) inside MQL5 from two iMA LWMA primitives: a half-period LWMA(17) and a full-period LWMA(34). The differenced series 2*WMA_half - WMA_full is then smoothed by a hand-rolled LWMA over sqrt(34) = 5 samples in HullMAAt(). The slope of that HMA is checked between shift=1 and shift=2 on the closed bar, and the close must also sit on the correct side of the HMA. The trend filter is then gated by a 14-period RSI whose OB/OS bands are themselves ATR-adaptive: when ATR-percentile climbs above 0.75 the bands widen by 5 points (RSI_BASE_OB 70 → 75, RSI_BASE_OS 30 → 25) to absorb fakeouts in high-vol bursts, and when ATR-percentile drops below 0.25 the bands tighten by the same amount so entries can fire earlier in quiet regimes. The third gate is an inlined Heikin-Ashi colour test — HAAt() walks 15 bars oldest-to-newest from CopyOpen/High/Low/Close and seeds HA_Open with the OHLC4 average, then iterates (prev_HA_Open + prev_HA_Close)/2. A long entry requires HMA rising, close > HMA, RSI between 50 and the adaptive OB, and the latest HA candle green; shorts mirror.

Regime is a composite of ADX(14), an ATR(50) percentile rank across the lookback, and Bollinger Bands(20, 2.0) width. RegimeClassify() returns seven states: StrongTrend (ADX ≥ 25 and ATR-pctile ≥ 0.6), WeakTrend (ADX ≥ 20 and pctile ≥ 0.4), Range, Breakout (BB expanding with pctile ≥ 0.7), Compress (pctile ≤ 0.20), Expand (BB expanding with pctile ≥ 0.5), and Choppy (ADX < 15 or buffer failure). Compress and Choppy both hard-block entries in SignalComposite() so the EA doesn't fade a sleeping market. The InpADXMin = 20 input feeds both the WeakTrend threshold and the StrongTrend gate (which adds 5 points on top).

Confirmation is a three-pronged test in ConfirmOK(). First, an HTF EMA(50) must agree with the trade direction — long only if close[1] >= emaHTF[0], short only if below. The HTF selection is InpHTFTimeframe = 8 which the MapTimeframeInt() helper maps via the default branch to PERIOD_H1, so out of the box the EA takes its higher-timeframe cue from the H1 trend. Second, the SL/TP distance must be feasible: it pre-checks slDist = 3.0 * ATR(14) and tpDist = slDist * InpMinRR(1.5) = 4.5 * ATR(14), so anything that would push either below the broker stops-level is filtered. Third, the server clock must be inside the 7-21 hour window defined by SESSION_START_HOUR and SESSION_END_HOUR. That keeps the EA trading the London-through-New-York gold session and stepping aside for the Asia lull.

Risk and sizing live in CalcLot() and TryEntry(). With InpRiskPercent = 0.5 the EA risks 0.5% of effective capital per trade, where effective capital is min(InpCapitalCapAmount, accountEquity) whenever the cap input is set non-zero; with the default InpCapitalCapAmount = 0.0 the cap is disabled and effective capital is just account equity. The $50 floor (InpCapitalCapFloor) still acts as a hard block on new entries even when the cap is off — once equity drops under $50 the EA prints a single CAP-BELOW-FLOOR message and stops opening new positions for the session. The actual lot is computed from (riskMoney / tickVal) * (tickSize / slDist), then floor-rounded to the symbol's lot step and clamped to LotsMin/LotsMax. OrderCalcMargin() is called before the send so a margin shortfall short-circuits the order without burning a request to the broker. Daily loss is a real-money check too: TodayRealizedPnL() walks the deal history from DayStamp(TimeCurrent()) to TimeCurrent() and adds profit + swap + commission only for deals with our magic, and once the absolute value reaches 3% of effective capital the entry pipeline is gated for the rest of the day.

Exits are stacked. The most aggressive is opposite-signal: ManageInTrade() calls SignalComposite(RegimeClassify()) every tick and closes the position if the signal flips. The most patient is time: MAX_BARS_IN_TRADE = 288 means a position is force-closed after ~24 hours on the M5 chart even if everything else is quiet. In between sits a 50% partial close at the 1R mark, a break-even move to open ± 2 points at 1R, and a forward-only ATR trail at 2.5 * ATR(14) that only ever tightens above (for longs) the entry price. SL is sent to the broker at 3x ATR(14) and TP at 4.5x ATR(14) for a fixed 1:1.5 R:R, but the in-trade logic actually runs the show — by the time the broker-side SL would be hit, the EA has usually either trailed past it or already closed on an opposite signal.

Scaling is the part most EX18s leave off. PyramidAllowed() permits up to MAX_PYRAMID_LEVELS = 3 positions on the same side, but only when each existing same-side position is already in profit by at least PYRAMID_ATR_SPACING = 1.5 * ATR(14) from its open. This makes pyramids a 'ride the winner' feature rather than a martingale: a 1.5-ATR move on gold at typical M5 ranges is several hours of trend, so the second and third adds only fire on genuine continuation. The MAX_POSITIONS = 5 global ceiling still caps total exposure, and MAX_DAILY_TRADES = 10 keeps one bad session from turning into a hundred-order blow-up.

The OnTester formula is the family standard — (netProfit * profitFactor) / (1 + balanceDrawdownPct), with a 30-trade floor. That criterion is what PipsGrowth's optimizer maximizes, so when back-testing this EA the genetic pass will tend to prefer parameter sets that combine high net return with low drawdown rather than raw profit. On a slow broker, expect 1-2 trades per session on quiet days and 4-6 on volatile ones; the MAX_DAILY_TRADES = 10 ceiling is almost never the binding constraint for gold M5.