Pipsgrowth EX18032 TrendFollow

Pipsgrowth EX18032 is a v2.00 trend-following EA built around a single classical core — a fast/slow EMA crossover on the last closed bar — wrapped in an unusually strict five-filter confirmation gate. Where many trend EAs vote between two or three momentum conditions, EX18032 demands that every confirm agree before a position opens. The cross itself is the standard 12/26 EMA pair (InpFastEMA=12, InpSlowEMA=26) evaluated on bar[1] against bar[2] inside ComputeSignal(), so the engine only sees fully closed bars and never trades on a forming candle. That alone removes the most common entry-timing complaint in M5 EAs: signals that fire on the open and then reverse before the close.

The first confirm is the trend EMA(50) — InpTrendEMA on the same chart. A long cross is only valid if the fast EMA(12) is above the trend EMA(50) on bar[1]; a short cross is only valid below. This blocks the classic EMA-cross failure pattern where the fast line jumps above a still-downtrending slow band and immediately snaps back.

The second confirm is an inlined Heikin-Ashi color check on bar[1]: haC1 = (O+H+L+C)/4 of the closed bar versus haO1 = (O[2]+C[2])/2 of the previous raw candle. Bullish HA color (haC1 > haO1) is required to take a long; bearish HA color is required to take a short. This is the single-bar HA form, not a multi-bar walk, so it gives a fast momentum pulse rather than a smoothed trend verdict.

The third confirm is an inlined SuperTrend side computed on the same chart with a 3.0x ATR(14) multiplier. SuperTrendSide() walks back through H/L/C and the ATR buffer, maintains final-up and final-down levels, and returns +1 (price closed above the upper band) or -1 (price closed below the lower band) on bar[1]. A long needs SuperTrendSide >= 0; a short needs SuperTrendSide <= 0. Because it is rebuilt from iATR plus CopyHigh/CopyLow rather than loaded as iCustom, it leaves no extra handle behind and works on any broker that supplies clean OHLC.

The fourth confirm is the higher-timeframe agreement test. The engine computes a single EMA(50) on a user-selected HTF (InpHTF=8 falls through MapTimeframeInt's switch and resolves to PERIOD_H1, the documented fallback). The previous HTF close is compared to that EMA: long trades require the HTF close above the H1 EMA(50), shorts below. This pulls the M5 entry into line with the hourly bias without forcing a multi-timeframe indicator stack.

The fifth layer is a seven-state regime classifier inside ComputeRegime(). It reads ADX(14), ATR(14) over a 100-bar percentile rank, and Bollinger Bands(20, 2.0) width as a percentage of price. The states are STRONG_TREND (ADX >= InpADXMinTrend=20 and ATR-percentile >= 60), WEAK_TREND (ADX >= 14, the 0.7x scaled floor), EXPAND (ATR-percentile >= 85), BREAKOUT (ATR-percentile >= 70 with BB-width% > 0.0015), COMPRESS (ATR-percentile <= 20), RANGE (ATR-percentile <= 40), and CHOPPY (the residual bucket). CHOPPY is hard-blocked in OkToTrade(); RANGE and CHOPPY together knock 25 points off the confidence score. STRONG_TREND adds 15; WEAK_TREND adds 5. The base score is 70 and it never exceeds 100.

The position manager runs every tick. ManagePositions() walks PositionsTotal() from the most recent position backward, filtering by magic 22218032 and the chart symbol. For each open position it checks four exit conditions in order: opposing EMA cross (close at bid/ask with reason 'OPP_SIG'), regime collapse to CHOPPY, or bars-in-trade hitting CONST_MaxBars=288 (roughly 24 hours on M5). When none of those fire, the manager runs the standard ratchet: at +1R (InpBEtriggerR=1.0) the stop is pulled to break-even on entry plus the broker stops-level offset; at the same +1R trigger 50% of the position is closed at the +1R mark (CONST_PartialFrac=0.5) if the residual still meets the symbol's minimum lot and step; thereafter an ATR trail with InpTrailAtrMult=2.5 advances the stop behind the bid or ask but never retreats below the break-even level.

The no-trade gate in OkToTrade() is dense. It rejects when the live spread exceeds InpMaxSpreadPt=50 points, when the time falls outside the 08:00-20:00 server window defined by InpSessionStart and InpSessionEnd, when the per-magic cooldown is active, when the live open count has reached InpMaxOpen=3, when EffectiveCapital() drops below the $50 floor (InpCapFloor), or when the realized daily loss has hit -3% of effective capital or the realized weekly loss has hit -6%. EffectiveCapital() returns min(InpCapAmount, equity) plus today's realized PnL when the cap is enabled, or pure equity when InpCapAmount=0 (the default-off state). The cooldown is engaged by UpdateRealizedPnL() when it counts three or more losses in the current day — CONST_CooldownLoss=3 — and holds for CONST_CooldownHours=4 hours, after which RollDayIfNeeded() clears the streak on the next day rollover.

Sizing comes from a clean lot formula: lots = (cap * InpRiskPercent/100) / (slDist * tickValue/tickSize * point). The default risk per trade is 0.5% (InpRiskPercent=0.5), the SL distance is 1.8x ATR(14) (InpSLatrMult=1.8), and the TP is 3.6x ATR(14) (InpTPatrMult=3.6) — a 1:2 reward-to-risk that passes the CONST_RRmin=1.5 pre-trade gate. Lots are clamped to the symbol's min/max volume and rounded to the volume step. Margin is pre-checked with OrderCalcMargin before the order is sent, and TryEntry retries once on a REQUOTE or TIMEOUT retcode. Close and modify helpers TryClose_EX18032, TryClosePartial_EX18032, and TryModify_EX18032 each run up to three attempts with 200ms or 100ms sleeps on REQUOTE/TIMEOUT/PRICE_OFF/PRICE_CHANGED before giving up.

The EA defaults to InpDryRun=true, which prints intended entries and exits to the journal without sending any orders — flip it to false in the inputs panel before going live. The InpKillSwitch input halts all OnTick processing immediately. The MT5 strategy tester hook OnTester() returns (netProfit * profitFactor) / (1 + balanceDrawdownPercent) as a single composite fitness value, with a hard floor of 30 trades before the score is computed; that is the value Strategy Tester will use when running a genetic optimization pass.

The design choice that defines EX18032 is the strictness of the gate. Most M5 trend EAs of this family use either a vote threshold (three-of-four, five-of-five at 60%) or a softer confirm list. EX18032 takes the opposite approach: every one of the five confirmations must align, the regime must not be CHOPPY, the spread must be inside 50 points, and the session must be open. The expected behavior in backtest is therefore a low trade frequency with a higher per-trade quality — extended sequences of dry bars interrupted by clean directional moves. On XAUUSD M5 with a $100 minimum account, that profile suits a trader who would rather watch a few well-filtered signals per week than churn through ten mediocre ones per day.

What to expect when running it: a $100 account with the default 0.5% risk loses at most $3 on a worst day before the EA pauses itself for the rest of the session, and at most $6 across a full week. The capital cap is off by default, so EffectiveCapital() returns the full account equity; turn InpCapAmount on (e.g. set it to 300) if you want the EA to size positions against a fixed allocation rather than your full balance. A standard XAUUSD ECN or raw-spread account is required — the 50-point spread filter is too loose for wide ECN accounts on quiet sessions, and a sub-20-point spread environment is necessary for the partial-TP-at-1R mechanic to fill cleanly.