Pipsgrowth EX18016 TrendFollow

Pipsgrowth EX18016 stacks two trend primitives — a hand-coded Hull Moving Average and a hand-coded SuperTrend — and only fires when both agree, the higher-timeframe direction supports them, and the regime is one of four expansion- or trend-flavoured states. The result is a USDJPY-tuned M15 trend-follower that does not chase mean-reversion, fade breakouts, or act on noise. It is engineered for traders who want a clean trend model with multi-timeframe agreement, explicit regime gating, and a conservative scaling rule that most default to single-position.

The signal layer is built from scratch inside the EA, not delegated to the iCustom handle. The Hull MA fast primitive uses two linearly weighted moving averages with period n/2 and n, then takes a third LWMA over the difference series with period sqrt(n). The HullMAAt() function pulls both underlying WMA buffers, builds the 2*WMA(n/2) − WMA(n) differenced array, and applies a triangular kernel of weights over the sqrt(n) samples. The result is a Hull curve with minimal lag that still filters out fast wiggles. The fast Hull period defaults to 20. A slope check (current value vs prior bar) decides direction.

The SuperTrend is also inlined. The SuperTrendAt() function walks backward over a window of 50 bars, holding an upper band, a lower band, and a direction flag. At each bar it computes the basic upper/lower from HL2 ± ATR*InpSTMultiplier, then tightens the bands using the classic carry-forward rule (raise lower only when the prior close is above the previous upper; lower upper only when the prior close is below the previous lower). Direction flips when the close breaks the active band. The result is the same SuperTrend line most traders are familiar with, but recomputed from inlined ATR data so the EA owns the math end-to-end. Default ATR period is 10, default multiplier is 3.0.

The signal fires only when three things agree: the fast Hull is sloping up (or down), the inlined SuperTrend direction is +1 (or −1), and the closed-bar close is on the correct side of the Hull. This triple-lock eliminates most single-indicator whipsaws. SignalComposite() returns +1 for a buy trigger, −1 for a sell trigger, and 0 otherwise. The function explicitly suppresses signals in regime codes 2 (Range), 4 (Compress), and 6 (Choppy) — the three states where a trend model has no edge.

RegimeClassify() runs a 7-state classification on every cycle. ADX (period 14) above 25 with ATR-percentile (50-bar lookback) at or above 60% is StrongTrend. ADX above 20 with percentile at or above 40% is WeakTrend. Bollinger width expansion with percentile at or above 70% is Breakout. Bollinger width expansion with percentile at or above 50% is Expand. ATR-percentile at or below 20% is Compress. ADX below 15 is Choppy. Everything else is Range. A 4-bit mask (REGIME_GATE_MASK = 0x4B) restricts new entries to StrongTrend, WeakTrend, Breakout, and Expand. Compress, Range, and Choppy are blocked at the regime layer, and RegimeAllowed() enforces it again as a redundant belt-and-suspenders in NoTradeBlock().

ConfirmOK() adds a higher-timeframe filter. The HTF Hull MA (period 50, default H1) must slope the same way as the fast Hull on M15, and the closed-bar close must sit on the correct side of the HTF EMA(50). Both conditions are mandatory. A R/R feasibility check is included so the EA only fires when SL distance and TP distance (1.5×SL) are both positive and the implied reward-to-risk clears the InpMinRR floor of 1.5. A server-time session window is also enforced (default 0–23, full day, with the EA's IsWeekendClose() closing every position on Friday at 22:00 server time and refusing to open new ones across Saturday and Sunday).

Risk and sizing follow a standard 0.5%-of-equity model with a 0.01-lot fixed fallback when the account is too small for proper risk math. The capital cap is built in: EffectiveCapital() returns min(InpCapitalCapAmount, account_equity), and a hard floor of 50 USD blocks new entries once effective capital drops below the floor. The daily loss limit is 3% of effective capital, calculated from realized P&L since the day's midnight stamp. The default capital cap amount is 0.0, which disables the cap and uses raw equity — set it above 0 to throttle exposure to a fixed dollar figure.

The pyramid stack is the most unusual design choice. InpPyramidEnabled defaults to false, so out of the box the EA holds a maximum of one position per side. When enabled, PyramidAllowed() enforces a triple-gated rule: the new add requires (1) every same-side position to already be in profit by at least 1.0 ATR, (2) every same-side position's stop to be secured (SL above open for longs, below open for shorts), and (3) the cap of InpPyramidMaxLevels (default 3) per side not to be breached. The cap is then also clamped to MAX_POSITIONS (5). This is a strict, conservative add — not a martingale, not averaging into losers.

Trade management runs every tick from ManageInTrade(). The TryClose_EX18016, TryClosePartial_EX18016, and TryModify_EX18016 wrappers each retry up to three times on requote, timeout, price-off, or price-changed errors, with 200ms and 100ms sleeps between attempts. A time-based exit closes positions that have been open longer than MAX_BARS_IN_TRADE (96 bars, ~24 hours on M15). An opposite-signal exit closes longs when the live signal flips negative (and vice versa) — this preserves the source's trend-flip-exit behaviour. A 50% partial close fires once the position is in profit by at least 1R, provided the remaining volume is at least 1.5× the symbol's minimum lot. After the position is 1R in profit, the stop is moved to break-even plus a 2-point buffer. From that point, an ATR trail at 2.5×ATR ratchets the stop forward in 0.5×ATR steps. When the regime is WeakTrend, the trail distance is halved (1.25×ATR) so weak-trend winners lock in faster.

The OnTester() criterion is (net * profit_factor) / (1 + balance_drawdown) with a 30-trade minimum. This rewards strategies that combine profitability with low drawdown relative to the absolute profit — a much harsher selector than raw net profit, and one that favours a smooth equity curve over a high-amplitude one.

In a backtest on USDJPY M15 over a multi-year window with default settings, expect a low-frequency, low-trade-count profile with most entries occurring during London and New York overlap hours and a sizeable number of entries being filtered out by the Choppy and Compress regime gates. Default InpDryRun is true, so the EA logs every intended trade to the journal without sending orders — set it to false before going live.