Pipsgrowth EX16038 Trend

EX16038 is a strict, multi-filter trend-rider for XAUUSD on M5 that demands five independent confirmations to align before any ticket hits the wire. The EA reads two EMA pairs in parallel — a 21/50 EMA on the working timeframe and a 21/50 EMA on a user-selected higher timeframe (defaults to H1 via the HigherTF_EMA=8 token that MapTimeframeInt translates to PERIOD_H1) — and only flags a buy when fast sits above slow on both timeframes and the corresponding sell setup is mirrored. That double-EMA alignment is the regime classifier: the EA treats a misaligned higher timeframe as a hard veto against taking the trade.

The rest of the EvaluateSignalOnClosedBar() routine layers four more confirmations on top of the EMA cross. ADX(14) must print at or above ADXMin=25, with +DI > -DI for buys and -DI > +DI for sells — that branch alone adds one point to whichever side's running score. RSI(14) must clear RSI_BuyMin=55 on the buy side or sink to RSI_SellMax=45 on the sell side, again adding one scoring point. The GetDonchian() helper then walks the 55 most recent closed bars, computes the highest high and lowest low, and accepts a buy only when the closed-bar close prints above the upper band or a sell when it prints below the lower band — the classic Donchian breakout filter, used here to suppress mean-reversion entries inside the channel. Finally, ATR(14) must clear ATRMinPoints=20 to add the volatility-floor point. The combined score can reach six, and the EA requires scoreBuy >= 5 with scoreSell <= 2 before a buy fires (mirrored for sells). The two conditions of OnlyTrendDirection=true (counter-trend signals disabled) and the EMA dual-timeframe alignment make this EA effectively a trend-only system with built-in mean-reversion suppression.

Entry execution lives in TryOpenDirection() and is gated by CanOpenNewEntry(). The gate runs four checks in order: the loss-streak pause timer (g_tradePauseUntil) must have elapsed, the total open position count must be below MaxOpenTrades=5, the current server hour must fall inside either London (8-17) or New York (13-22) when the session filter is on, and the spread must clear the dynamic cap. The spread cap is itself a moving target: UpdateSpreadStats() writes the current spread into a 512-element ring buffer, GetMedianSpread() runs an insertion sort and returns the median of the last SpreadMedianWindowSize=200 samples, and CheckSpreadCap() compares the live spread against SpreadCapMedianMultiplier=3.0 * median — the EA will not fire a new entry when the live spread is more than three times the recent median, which is its primary protection against broker-side spread spikes during news.

Position sizing runs through CalcLotsFromRisk(). When UseRiskPercentSizing=false (the default), the EA simply uses FixedLotSize=0.10. When the user flips the toggle on, the routine divides RiskPercentPerTrade/100 * balance by the per-lot risk measured in ticks — ticksInSL * g_tickValue where ticksInSL = slPoints * _Point / g_tickSize — and the resulting lots are normalized through ClampVolume() against the symbol's SYMBOL_VOLUME_STEP, _MIN, and _MAX. The risk math falls back to the fixed lot if any of the cached _tickSize, _tickValue, or balance inputs are non-positive, so a misconfigured broker or thin symbol won't crash the EA — it just reverts to the fixed lot.

SL and TP are fixed in points: StopLossPoints=300 and TakeProfitPoints=600 produce a 30/60 pip R:R on a 5-digit XAUUSD broker, with EnforceStopLevel=true causing RespectStopDistanceOK() to discard either stop if it falls inside the broker's SYMBOL_TRADE_STOPS_LEVEL. The EA ships with UseECNSafeOrderSend=true, which makes SendOrder() fire the market order without SL/TP first and then ModifySLTPByTicket() apply both levels after a short polling loop (10 attempts at 50ms) finds the new position — a workaround for brokers that reject distance-violating SL/TP on the initial fill. The same EnforceFreezeLevel=true flag also causes WithinFreezeLevel() to skip modify attempts within the broker's SYMBOL_TRADE_FREEZE_LEVEL, so trailing logic won't fight the broker near expiry.

The Dynamic Lock Profit ratchet is the EA's signature position-management feature. MaintainPositionsDLP() walks every open magic-symbol position once per tick and calls ApplyDynamicLockProfitToPosition(). The function reads the live bid (for buys) or ask (for sells), computes movedPts, takes steps = floor(movedPts / 150), and pushes the new SL to entry + (steps * 150 - 20) * g_point for buys or entry - (steps * 150 - 20) * g_point for sells. That is, every 150 points of favourable move the EA pulls the stop closer by 150 points minus a 20-point buffer. The ratchet only ever tightens — curSL==0 || targetSL > curSL for buys, mirrored for sells — and it respects both the freeze and stop levels before calling ModifySLTPByTicket(). At default settings the EA will start protecting profits once the trade is 150 points in the money and ratchet in 130-point increments thereafter.

Additions are also gated. GateAdditionalOK() checks that the same-direction count is zero (the initial position) before allowing the trade, and when AllowOnlyProfitableAdditions=true it then calls IsAllSameDirProfitableAtLeast() to verify that every same-direction open position has moved at least MinProfitPerTradeToAdd=5.0 points from entry. The build comment pattern is explicit: the first order tagged Initial BUY or Initial SELL, with subsequent entries tagged Add#1, Add#2, etc., and the g_nextAddIndexBuy/g_nextAddIndexSell counters increment each time.

Loss-streak handling lives in OnTradeTransaction(). Every DEAL_ADD that is a DEAL_ENTRY_OUT for this EA's magic/symbol pulls the profit+swap+commission P&L, and if it is negative it increments g_lossStreakCount and adds the absolute loss to g_lossStreakAmt. With PauseOnConsecutiveLosses=false and PauseOnLossAmount=false (both off by default) the EA never sets a pause — but if the user flips either toggle on, hitting PauseConsecutiveLossesCount=3 consecutive losses pushes g_tradePauseUntil 60 minutes into the future, and accumulated g_lossStreakAmt >= PauseLossAmount=100.0 USD does the same. Any winning exit zeroes both counters. The pause is checked at the top of CanOpenNewEntry() and is the only global entry veto.

Three TryClose_EX16038 / TryClosePartial_EX16038 / TryModify_EX16038 helper functions sit at the very bottom of the file as a 3-attempt REQUOTE / TIMEOUT / PRICE_OFF / PRICE_CHANGED retry wrapper around the standard CTrade class. They are defined but never called — every live close and modify path goes through SendOrder(), ModifySLTPByTicket(), or the polling loop inside SendOrder() with its own retry behaviour, so these helpers are dead code from the runtime perspective. They look like a leftover scaffold from a refactor that left the dedicated retry path behind.

The header advertises all twelve architecture layers — REGIME, SIGNAL, ENTRY, CONFIRM, NO-TRADE, CAPITAL CAP, RISK, SIZING, MANAGE, EXIT, SCALING, OnTester. The first nine are visibly wired: the dual-EMA HTF alignment is the regime layer, EvaluateSignalOnClosedBar() is the signal/confirm combo, the CanOpenNewEntry gate is the no-trade layer, the fixed-lot / risk-percent paths are sizing, the DLP ratchet is the manage layer, and the fixed SL/TP with the close-on-fill polling is the exit path. The CAPITAL CAP layer — balance drawdown cap, equity stop, daily-loss limit — is not implemented: there is no balance ceiling, no equity stop, no daily or weekly loss limit, and no max-trades-per-day. EXIT happens entirely through SL/TP and the DLP ratchet — there is no reverse-signal close, no time-based exit, no break-even function. OnTester is described in the header but not present in the source. The EA also has a MapTimeframeInt() function declared four times in identical form, which is harmless because the linker dedupes it but does add to the file's noise.

For a first live run, default settings on a $100 ECN account will produce one 0.10 lot entry per closed bar inside the London or NY window, holding up to 5 positions in the same direction with the ratchet stepping in at every 150 points of favourable move. Spread spikes outside the rolling 3x median will gate entries but not exit live positions, so a sudden spread widening while a position is open will not trigger an early close. The Capital Cap layer being absent is the most material gap to be aware of: a long sequence of losing trades will keep opening new positions until MaxOpenTrades=5 is hit, with the only interruption being a manual pause toggle flip or a session boundary.