Pipsgrowth EX18034 TrendFollow

EX18034 is a four-filter trend consensus engine. Rather than betting the trade on a single moving-average cross, it asks four independent questions on the last closed M5 bar and only acts when the answers line up. The four filters are: where the close sits relative to a 50-period EMA on the current timeframe, the sign of an inlined Hull MA(20) slope, the side of an inlined SuperTrend(22, 3.0×ATR), and the colour of the most recent Heikin-Ashi candle. Each filter gets one vote, and TrendSignal() returns a directional bias only when at least three of the four agree and the fourth is silent. Confidence is then reported as the vote count times twenty-five, so a 3-vote agreement prints 75 and a unanimous 4-vote print returns 100; the value is informational because the gate is the 3-of-4 rule, not the integer.

The Hull MA(20) is rebuilt in code from two LWMA handles rather than fetched from a built-in Hull indicator. HullValue() pulls the full-period LWMA and the half-period LWMA, computes two-times-half minus full, then runs a final LWMA of length sqrt(20)≈4 over that difference — the textbook Hull reconstruction. HullSlope() compares the value at shift=1 with shift=2 and returns +1, -1, or 0. The SuperTrend is also inlined. SuperTrendSide() holds the previous band and direction in module-level state (st_prev_dir, st_prev_upper, st_prev_lower) so the bands can ratchet in the standard SuperTrend way: a close above the prior upper band flips the side to +1, a close below the prior lower band flips to -1, and the current band is whichever of the local upper/lower or the carried value is tighter given the previous close. Heikin-Ashi color is computed by recursively walking back two bars — the HA-Open of the last closed bar is the average of the previous HA-Open and HA-Close, and HA-Close is the mean of the underlying OHLC. No third-party indicator handles are required for the signal stack.

The regime gate that sits on top of the consensus is a five-state classifier, not the seven-state version seen on some siblings. RegimeGet() reads ATR(14), ADX(14), and Bollinger Bands(20, 2.0) and computes the ATR-percentile over a 100-bar window (InpAtrLookback=100) and a parallel BB-width percentile over the same lookback. The five states are: REG_BREAKOUT when ADX is at or above InpAdxTrendMin (20) and both the ATR and BB-width percentiles sit at or above the 70th percentile, REG_STRONG_TREND for ADX above 20 without the volatility-pop confirmation, REG_WEAK_TREND for ADX between 18 and 20, REG_RANGE for low-volatility compression at the 30th percentile or below on both ATR and BB-width, and REG_CHOPPY as the residual catch-all. The 100-bar percentile is a longer memory than the 50-bar version used elsewhere in the family, which makes the regime label slower to flip and less prone to whipsawing around news spikes.

The confirm step applies after the signal fires. Confirm() reads the H1 EMA(50) and demands that the H1 close sits on the correct side of it, refuses to act when ADX is below 18 even if the four-vote consensus passed, and rejects trades when the configured reward-to-risk is below 1.0. There is no built-in higher-timeframe selector input — the HTF is hard-coded to PERIOD_H1. The 1.5 R:R default feeds straight into the position builder, so the take-profit distance is 1.5× the stop distance.

Stop placement uses ATR(14) and a fixed 2.0× multiplier. SendOrder() also enforces the broker's stops-level and freeze-level floors via AdjustForStops(), pushing the SL and TP a few points further out if the broker's minimum would otherwise be violated. The take-profit is set 3.0×ATR from the entry (2.0×ATR stop times the 1.5 R:R ratio). All lots are sized off effective capital, which on a fresh run equals account equity but can be reined in by InpCapAmount when a non-zero value is set. The default cap of 0 means the cap is effectively disabled and full equity is used. CalcLots() multiplies effective capital by the 0.5% risk percent, divides by the per-lot loss implied by the ATR stop distance and the symbol's tick value, then normalizes to the broker's lot step. CalcLots returns zero if effective capital has fallen below InpCapFloor (50 by default), which is what blocks entries on a blown account.

In-trade management has three legs and runs every tick regardless of whether a new entry is being considered. Once price has moved one R in the favourable direction, the stop is ratcheted to entry plus two points minus the broker stop floor, locking in a scratch-free break-even. The 50% partial close fires the moment the bid/ask touches open plus 1.5×R, halving the position and letting the remaining half run. The trailing stop is a 2.5×ATR forward-only ratchet that walks on every tick and only ever tightens — it never moves backwards. If the four-vote consensus flips against the open position, the entire position is closed at market through TryClose_EX18034(), which loops up to three attempts with 200 ms sleeps on requote, timeout, price-off, and price-changed retcodes. There is no time-based exit, no daily-trade cap, and no weekly loss limit on this build — the only account-level kill is the per-magic 3% daily realised PnL cap and the global InpKillSwitch input.

Pyramid add-ons are compiled into the binary but the default input is false, so out of the box EX18034 will not stack positions. If you flip InpPyramidEnabled to true, PyramidAllowed() will gate add-on entries behind regime not being CHOPPY or RANGE, a fresh bar having formed since the last entry, total position count being below the lesser of five and InpMaxOpen, and the existing same-direction position already in profit. The OnTester criterion is the standard (net × profit-factor) / (1 + relative drawdown) with a 30-trade floor, so the strategy optimizer will favour runs that combine profitability with a smooth equity curve rather than a single lucky streak.

In dry-run mode (the default), the EA prints what it would have done without sending orders. Switch InpDryRun off to go live. Order comments are tagged with the magic 22218034 and the string Psgrowth.com Expert_18034 so backtests and live journals can be filtered by comment. The header lists XAUUSD M5 as the suggested deployment but every parameter is symbol-agnostic — point out the EA at any XAU/USD, XAG/USD, or major pair that your broker offers, on any timeframe, and the same logic will run.