Pipsgrowth EX18083 TrendFollow

EX18083 is a single-file MT5 trend follower that builds its signal around a Hull Moving Average whose own period is allowed to breathe. The author calls it "JMA Adaptive" in the brief, but the implementation in GetSignal() is a classical HMA: HMA(period) = WMA(2*WMA(period/2) − WMA(period), sqrt(period)). The adaptive twist is that the period itself is rewritten every closed bar from a 30-bar ATR(30) ratio. The EA copies the last InpBasePeriod (30) ATR values, computes the mean, divides the mean by the most recent value, clamps the result to the [0.5, 2.0] band, and multiplies InpBasePeriod by that factor — yielding a final period in the 15–60 range depending on whether the market is hot or quiet. The HMA is then rebuilt from scratch on every bar inside CalculateHMA(), since the period is variable the code cannot cache a single iCustom buffer. Each rebuild reads roughly period + sqrt(period) + 10 closes and runs two WMA passes and one outer WMA over the resulting difference series.

With the HMA in hand, GetSignal() looks at the HMA on the just-closed bar (shift 1) and the bar before that (shift 2). A simple slope test — hma1 > hma2 for an uptrend, hma1 < hma2 for a downtrend — defines direction. A distance gate sits on top: the close of the trigger bar must lie on the correct side of the HMA, and MathAbs(close1 - hma1) must exceed current_atr * InpSignalAtrFactor (0.25 by default). The distance threshold is therefore a fraction of the current ATR, not a fixed pip count, which lets the EA accept wider pullbacks in volatile regimes and demand more separation in calm ones. If both slope and distance agree, the function returns +1 for long or −1 for short. If anything fails it returns 0 and the EA does nothing that bar. There is no higher-timeframe confirmation, no ADX filter, no news gate, and no regime classifier in the source — InpUseFlatFilter is exposed in the input list but GetSignal() never reads it, so toggling it on or off has no effect on entries.

OnTick() is a thin, single-position manager built around a closed-bar gate. The first thing it does is call ManageProfitLock() every tick, so trailing is continuous even when the bar is not yet closed. Then the time filter (InpUseTimeFilter = true, 08:00–20:00 server by default) — if InpCloseAtEnd is enabled, the EA flattens the basket at the end of the window; otherwise it simply stops opening new trades until the next session. A static last_bar variable gates the rest of the pipeline to one decision per M5 bar, after which GetSignal() is called. If the signal is non-zero, the EA hands the direction to CheckPyramidSafety(). That function refuses to add against an opposite-side position, refuses to add if the existing same-side position is not already in profit by at least InpPyramidMinProfit (100 points of secured profit, measured from open to current SL), and refuses if InpMaxPyramidTrades (3) is already reached. If the EA is running with InpUsePyramid=false, it acts as a strict single-position system and only opens the first trade in a series. The new trade is sized by GetLotSize(InpStopLoss), which prefers 1% balance risk and falls back to InpFixedLot (0.01) when InpRiskPercent is 0; the result is floored to the symbol's LotsStep, clamped to LotsMin/LotsMax, and capped at InpMaxLotSize (10.0). SL is InpStopLoss (500 points) and TP is InpTakeProfit (1500 points), giving a fixed 1:3 reward-to-risk on every entry — there is no ATR-scaling on the exits, so the structure of the trade is constant regardless of volatility.

ManageProfitLock() is a one-way ratchet that never loosens a stop. For a long, it computes the current profit in points; once that profit reaches InpLockTrigger (300 points), it sets a candidate SL at Bid − InpLockDistance * _Point (100 points behind price) and only sends a modify if the new SL is at least InpLockStep (50 points) above the current SL. For a short it mirrors the math. Because the trigger condition is strictly points_profit >= InpLockTrigger and the modify condition is strictly target_sl > sl + step (or target_sl < sl - step), the SL can only ever move in the profitable direction. The TP is left untouched by the ratchet, so the original 1500-point target stays in place until either TP or the ratcheted stop is hit. Closes go through TryClose_EX18083(), a 3-attempt wrapper that retries on TRADE_RETCODE_REQUOTE, TRADE_RETCODE_TIMEOUT, TRADE_RETCODE_PRICE_OFF, and TRADE_RETCODE_PRICE_CHANGED with a 200ms back-off; the corresponding modify wrapper TryModify_EX18083() uses a 100ms back-off.

The filling mode is hard-coded to ORDER_FILLING_FOK and SetDeviationInPoints(30) matches the InpSlippage input. The m_trade object is given magic 22218083, the comment "Psgrowth.com Expert_18083". There is no OnTester() block, no info panel, and no per-trade partial close — partials are wired through TryClosePartial_EX18083() but the EA never calls them. There is no drawdown breaker in the file: the only capital cap is the per-trade lot cap and the pyramid cap. Run it on a low-spread XAUUSD or EURUSD broker (the source assumes points-based pip math, so a 5-digit broker is required for the SL/TP/trailing values to read sensibly). On a $100 account the 1% risk default puts the per-trade stop at roughly the cash-equivalent of 100 points of pip-value, which on a 0.01-lot XAUUSD trade is small enough to survive the 500-point fixed SL without exhausting margin; raise InpMaxLotSize cautiously if you scale up the deposit. The most useful backtest probe is the InpSignalAtrFactor knob: lower it (0.15–0.20) for more entries in trends, raise it (0.30–0.40) for cleaner breakouts in noise.