Pipsgrowth EX18017 TrendFollow

EX18017 is a closed-bar trend follower built around two hand-rolled indicators that have to agree on direction before the EA will open a position. The first is a Hull Moving Average at period 55 — a longer setting than the typical 14 or 20, and computed in-line from two LWMA primitives (a WMA at n/2 and a WMA at n) differenced and smoothed again over √n samples. The second is a SuperTrend with a 10-bar ATR and a 3.0 multiplier, also computed in-line by walking backward through 50 bars of high/low/close data, maintaining upper and lower bands (HL2 ± mult × ATR), and flipping trend direction whenever a close crosses the active band. The 3.0 multiplier is wider than the conventional 2.0, so the band resists early whipsaws and the EA stays out of the first bars of a reversal. The 55-period Hull smooths through the same noise. Together they form a slow, deliberate composite that prefers persistence over sensitivity.

The entry vote is three-way. A long requires the Hull to be sloping up (HMA[1] > HMA[2]), the SuperTrend to be in direction +1, and the previous bar's close to sit above the Hull value. A short inverts all three. The signal is read on the closed bar, so a new position opens at the open of the next bar, not mid-tick. If any of the three legs disagrees, the EA returns 0 and does not enter.

A seven-state regime classifier runs in front of the signal. It uses ADX(14), a 50-bar ATR-percentile, and a Bollinger-width expand/contract check. StrongTrend is ADX ≥ 25 with percentile ≥ 60%. WeakTrend is ADX ≥ 20 with percentile ≥ 40%. Range is the catch-all low-conviction state. Breakout fires when Bollinger Bands are expanding and the ATR percentile is ≥ 70%. Compress is percentile ≤ 20%. Expand is expanding with percentile ≥ 50%. Choppy is ADX < 15. The pipeline blocks Choppy and Compress outright — those are the regimes where the wide SuperTrend would lag a real move and the EA would take a stop it shouldn't. Range, Expand, Breakout, StrongTrend, and WeakTrend are all permitted. The classifier recomputes every closed bar.

Confirmation runs after the signal fires. The EA pulls a 50-period EMA from the higher timeframe (default H1) and requires the entry bar's close to sit on the correct side of that line. A long signal with close[1] < H1-EMA(50) is rejected. A short signal with close[1] > H1-EMA(50) is rejected. That single H1 filter is the most important reduction in counter-trend losses on a 5-minute gold chart — it forces the EA to trade with the dominant cycle, not against it. The EA also checks a minimum reward-to-risk ratio of 1.5 using the ATR proxy: if the computed SL would not produce at least 1.5× its distance in TP, the trade is skipped before any sizing happens. The session window is hard-coded 07:00 to 21:00 server time.

Risk and sizing are percentage-based. The default is 0.5% of effective capital per trade. Effective capital is min(InpCapitalCapAmount, account equity) when the cap is enabled, otherwise raw equity, with a $50 floor below which new entries are blocked. Lot is computed from the per-trade risk budget divided by the per-lot risk at the actual SL distance in tick-value terms. If the proposed SL would land inside the broker's stops-level, the EA widens it by 10 points to clear the constraint; the TP widens in proportion. An OrderCalcMargin pre-check refuses the send if the required margin exceeds free margin.

Stop and target are ATR-derived. Initial stop is 3.0 × ATR(10). Target is 1.5 × that distance, so the trade opens at a fixed 1:1.5 reward-to-risk ratio. The 1.5 minimum-RR gate in confirmation is the same number, which means the EA will never open a position whose computed TP would be inside the minimum ratio.

In-trade management has three moving parts. When price moves 1R in favor, the stop is ratcheted to break-even plus a 2-point buffer (the buffer covers the spread and the commission so the BE move is not a guaranteed loss on the way out). After that, a 2.5× ATR trailing stop takes over and walks behind the position. At 1R, 50% of the position is closed — but only if the residual volume would still be at least 1.5× the broker's minimum lot, otherwise the partial is skipped to avoid leaving an un-closeable stub. Two hard exits complete the loop: a time exit at 288 bars (approximately 24 hours on M5), and an opposite-signal exit the moment the composite flips. Both run inside ManageInTrade(), which executes on every tick, not just on bar close.

Pyramid: the PyramidAllowed function exists and enforces two conditions — at most 3 same-side positions, and the existing position must be at least 2.0 × ATR in profit before a scale-in is allowed. The build has no input toggle to turn it on, so it sits in standby. A user who wants scaling would have to edit the source.

Execution and risk rails. The OnTester criterion is (net × profit factor) / (1 + balance drawdown), with a 30-trade minimum. Order sends retry once on requote, timeout, or price-off. Closes and partials run a 3-attempt loop with 200 ms between attempts. Modifications get 3 attempts at 100 ms intervals. Slippage is set to 30 points. The daily loss cap is 3% of effective capital. Max open positions is 5, max new entries per day is 10, and the spread is filtered by a rolling 50-bar average — the cap is the larger of 3× the average or (average + 20) points, so a sudden spread spike blocks one cycle, not the whole day. The EA ships with InpDryRun=true as the default, so it logs what it would have done without sending orders until you flip the switch.