Pipsgrowth EX18038 TrendFollow

Pipsgrowth EX18038 TrendFollow is an Ichimoku-driven pullback-and-trend EA built for XAUUSD on M5–H1 charts, with the 9/26/52 Tenkan/Kijun/Senkou-B triple native iIchimoku as its only signal source. Where the rest of the TrendFollow family uses inline EMA crosses, Hull, SuperTrend, KAMA, AMA, or Alligator primitives, EX18038 leaves the Ichimoku calculation to MetaTrader's compiled iIchimoku(_Symbol, _Period, InpTenkan, InpKijun, InpSenkou) handle and reads back the four buffers — Tenkan-sen (buffer 0), Kijun-sen (buffer 1), Senkou Span A (buffer 2), Senkou Span B (buffer 3) — once per new bar. That is the first design decision visible in the code: keep the indicator math where MetaTrader can vectorise it, then do the cross detection in IchimokuSignal() against last-closed-bar values only (shift 1) so no repaint ever leaks into the live book.

The IchimokuSignal() function returns +1 for long, −1 for short, 0 for no trade. A long requires tk[1] > kj[1] while tk[2] <= kj[2] (Tenkan crossing up through Kijun on the closed bar) AND close[1] > max(sA[1], sB[1]) (price closing above the top of the Kumo cloud). The short is the mirror: tk[1] < kj[1] while tk[2] >= kj[2] and close[1] < min(sA[1], sB[1]). The Kumo filter is what separates EX18038 from a plain moving-average cross — both lines have to agree on a freshly flipped cross AND price has to be on the correct side of the cloud at the close. The function also stamps a confidence of 80 in REG_STRONG_TREND and 60 in REG_WEAK_TREND, but that confidence is informational only and does not gate the entry — the regime block is enforced elsewhere.

Regime classification lives in ComputeRegime(). It reads ADX(14), ATR(14), and Bollinger Bands(20, 2.0) and combines them into six buckets: REG_STRONG_TREND when ADX ≥ 30, REG_WEAK_TREND when 20 ≤ ADX < 30, REG_COMPRESS when the Bollinger band-width ratio bwPct = (bbU[1] - bbL[1]) / bbM[1] falls below 0.015 (1.5%), REG_EXPAND when bwPct exceeds 0.045 (4.5%), REG_RANGE for the 1.5%–4.5% middle band, and REG_CHOPPY as a final override when ADX < 15. RegimeAllowsTrading() only returns true for STRONG_TREND and WEAK_TREND — EX18038 explicitly refuses to take entries in compress, range, expansion, or choppy conditions, and ManagePositions() force-closes any open position the moment the regime flips out of the two trend buckets. That is a meaningful design choice: the EA will exit a position that is still in profit if ADX drops back below 20, treating regime degradation as a higher-priority signal than the trailing stop.

Confirmation before entry runs through HTFTrend(), which reads the EMA from the higher-timeframe handle g_hHTFma = iMA(_Symbol, g_InpHTFTimeframe, InpHTF_EMA, 0, MODE_EMA, PRICE_CLOSE). The input InpHTFTimeframe defaults to 8 — that falls outside the 1–7 mapping in MapTimeframeInt() (1=M1, 2=M5, 3=M15, 4=M30, 5=H1, 6=H4, 7=D1), so the resolver falls into the default arm and the EA effectively pins itself to H1 in code, with EMA period 50. A long signal is killed if H1 close is below the H1 EMA, and a short is killed if H1 close is above it. The NoTradeReason() chain rejects on opposing HTF trend with the string HTF_TREND_OPPOSES.

The SL and TP are both ATR(14) multiples. InpSLATRMult = 2.0 puts the stop at twice ATR, and the TP is computed inline as slDist * 1.5, giving a 1:1.5 R:R structure. Both are clamped to (g_stopsLevel + 2) * g_point to respect the broker's minimum stop distance, and NormPrice() rounds to the symbol's digits.

Once a position is open, ManagePositions() runs on every tick and layers three protective mechanisms on top of the SL/TP. The break-even is a one-shot: when the floating R multiple reaches InpBETriggerR = 1.0, the stop is ratcheted to openPrice + stopsLevel*point (for longs) and never revisited. The ATR trail uses InpTrailATRMult = 2.5 — a long's new stop is bid - 2.5*atr; the trail only moves forward, never backward. The partial TP fires at InpPartialTP_R = 1.5R, closing exactly half the volume (NormVol(volume * 0.5)) provided half is at least the broker's g_volMin and strictly less than the full position, leaving the other half to run under the trailing stop. There is also a hard time exit at InpMaxBarsInTrade = 120 bars — ten hours of M5 bars — which closes the full position regardless of profit. On top of that, the opposite Ichimoku signal triggers a full close via ClosePartial() with the OPPOSITE_SIGNAL reason tag.

Risk and exposure controls run before entry. CalcLot() reads the effective capital — which is min(account_equity, InpCapitalCapAmount) if a capital cap is configured, otherwise just AccountInfoDouble(ACCOUNT_EQUITY) — and applies InpRiskPercent = 0.5% to size the position. Sizing uses the formula lot = (effCap * riskPct / 100) / (slDistance * tickValue/tickSize), then a pre-check against free margin via OrderCalcMargin(). If InpCapitalCapAmount is left at 0, the cap is disabled; if effective capital drops below InpCapitalCapFloor = 50, the EA blocks entries with the CAPITAL_BELOW_FLOOR reason. The daily loss guard is enforced by RollDayIfNeeded() — once realised P&L for the EA's magic number falls 3% of effective capital for the day, g_tradingDisabledToday is set and the next-bar signal is dropped with the DAILY_LOSS_LIMIT_HIT reason. There is no separate weekly cap; the cooldown is triggered after InpConsecLossTrigger = 3 consecutive losing closes, blocking new entries for InpCooldownMinutes = 30. InpMaxOpenTrades = 1 is unusually tight for the family — EX18038 runs a single-position book, and pyramid adds are compiled but switched off at the input level (InpPyramidMaxLevels = 0).

The no-trade filter chain in NoTradeReason() is one of the longer ones in the family and walks the same twelve checks as its siblings: kill switch, broker trade mode, capital floor, daily loss, cooldown, server-time session window (7–20 by default), manual news blackout (read from the global variable NEWS_BLOCK_UNTIL_<symbol> for InpNewsBlackoutMin minutes), Friday flat window (last 60 minutes before the weekly close), spread > 50 points, regime disallow, opposing HTF trend, stale quote (current bar time zero), stale bars (current bar more than 5× the period old), and CountMyPositions() >= InpMaxOpenTrades. Each rejection logs the specific reason tag to the journal, so a backtest can be post-mortemed by grepping the journal for BLOCKED lines.

Execution wraps the standard CTrade API with two retry layers. SendMarket() retries once on TRADE_RETCODE_REQUOTE, TRADE_RETCODE_PRICE_OFF, or TRADE_RETCODE_TIMEOUT after a g_sym.RefreshRates(). TryClose_EX18038, TryClosePartial_EX18038, and TryModify_EX18038 all run a 3-attempt loop with Sleep(200) (or 100 ms for modifies) between attempts, only retrying on requote/timeout/price-off/price-changed retcodes. Filling mode is auto-detected from the broker's SYMBOL_FILLING_MODE flag — FOK first, then IOC, then RETURN as the universal fallback — and the trade object's SetDeviationInPoints(20) allows 20 points of slippage on entry.

For the MetaTrader optimiser, OnTester() returns (net / maxDD) * profitFactor with a 10-trade floor. That criterion explicitly rewards strategies that produce profit per unit of drawdown, scaled by how well gross profit covers gross loss — a backtest that is net positive but has a brutal drawdown will score lower than a smoother equity curve with a similar net. Run any optimisation on XAUUSD M5 with at least one calendar quarter of data before reading the result as meaningful.

Two things to expect in live trading: first, EX18038 is a low-frequency EA. The Ichimoku TK cross on a single timeframe is not a frequent event, and the Kumo-cloud side filter plus the regime gate mean that on most days there will be zero entries, sometimes zero entries across an entire trading week. Second, because the trail only steps forward and break-even is a one-shot at exactly 1R, positions will sometimes sit through a 2–3R retracement before the trail catches up — that is the design, not a bug. If the operator needs faster exits, the right knob is InpTrailATRMult (lower = tighter trail), not the SL multiplier.