Pipsgrowth EX18056 TrendFollow

EX18056 is a single-symbol Hull-MA slope consensus engine that turns a two-line moving-average agreement into a trend-following bias. The engine keeps no manual pivot lists, no overnight position bookkeeping, and no chart-object drawing — the only state that survives between bars is the last signal bar time, which is the gate that prevents a single closed bar from firing twice in the same direction.

The signal itself is built in two layers. The first layer is an inline Hull Moving Average constructed from raw LWMA handles: a fast composite of period 20 (WMA(10) and WMA(20) blended with a sqrt(20) final smoother) and a slow composite of period 50 (WMA(25) and WMA(50) with a sqrt(50) smoother). The HullSignal() function reads both composites, takes the difference between the last closed value and the value InpHullLookback (default 3) closed bars earlier, and returns +1 only when both slopes point up, -1 only when both slopes point down, and 0 when the two Hulls disagree. Confidence is derived from the magnitude of the combined slope expressed in pips. There is no per-bar re-tinting and no candle pattern — slope agreement is the entire trigger.

The second layer is a regime gate. DetectRegime() samples ADX(14) and a 50-bar ATR percentile plus the growth ratio of Bollinger Band(20, 2) width, and assigns one of seven states: REG_STRONG_TREND (ADX ≥ 30 and ATR-pctile ≥ 60), REG_WEAK_TREND (ADX ≥ InpAdxMin=20 and ATR-pctile ≥ 40), REG_BREAKOUT (BB-width growth > 1.20 and ATR-pctile ≥ 65), REG_EXPAND (BB-width growth > 1.05 and ATR-pctile ≥ 50), REG_COMPRESS (BB-width growth < 0.85), REG_RANGE (ADX < 20 and ATR-pctile < 30), and REG_CHOPPY (everything else). The EntrySignal() function silently swallows the Hull signal whenever the regime is CHOPPY, RANGE, or COMPRESS, so the EA never tries to trade flat, mean-reverting, or volatility-collapsing tape.

Confirmation runs after the signal survives the regime filter. ConfirmEntry() reads the H1 EMA(50) value (the HTF handle is wired to g_InpHullHTF, which falls back to PERIOD_H1 because the input value 8 does not match any case in MapTimeframeInt) and rejects a long if the ask is below it, or a short if the bid is above it. It also blocks longs when RSI(14) is above 75 and shorts when RSI is below 25, and verifies that the live ATR-implied reward-to-risk ratio (InpAtrTPmult/InpAtrSLmult = 3.0/1.8 = 1.67) is at or above the InpMinRR=1.3 floor. Spread is rejected above 35 points. The HullSignal/EntrySignal/ConfirmEntry stack is what makes the engine feel selective on M5 gold without over-trading.

Position management is broken into three routines that run on every tick. ManagePositions() ratchets the stop with two parallel layers: a one-shot break-even that fires when unrealized profit reaches 1R and walks the stop to open plus 0.1R (offset), and a 2.5×ATR(14) forward-only trail that only ever tightens. ExitOnOpposite() closes any open position immediately when the live Hull signal flips sign, and ExitOnRegimeChange() force-closes everything whenever the regime becomes CHOPPY or RANGE, regardless of profit. There is no partial-close pathway, no time-stop, and no breakeven re-trigger — the BE is a one-shot per position.

Risk is bounded by four independent caps. Position sizing is 0.5% of effective capital per trade. A 3% daily realized-loss limit (counted off DEAL_PROFIT + DEAL_SWAP + DEAL_COMMISSION for this magic) and a 6% weekly realized-loss limit block new entries when hit. Max concurrent positions is 3, and max concurrent on this symbol is also 3, so the engine never crowds a single name. The optional capital cap (InpCapitalCapAmount=0 means off by default) shrinks the working capital to the smaller of the cap and account equity, and a $50 floor refuses to open new positions when effective capital falls below it. OrderCalcMargin is called before every send, so a margin shortfall is detected before the request reaches the broker.

Execution details that traders care about: InpDryRun is true by default — the EA prints the intended order instead of sending it until you flip the switch. Slippage is 30 points, the order comment is hardcoded to "Psgrowth.com Expert_18056", and a single transient retry is issued on REQUOTE or TIMEOUT. Pyramiding (InpPyramidEnabled) is off by default, but when enabled it requires the latest position to be at least 1.5×ATR in profit and caps total open positions at 3.

The time and session filters are conservative. Entries are blocked when the broker server hour is below 1 or at/above 23, when the day of week is Sunday or Saturday, or when Friday's server hour is 21 or later. Combined with the regime gate, the EA effectively concentrates its trading inside the most liquid part of the week and refuses to roll a position into the Sunday gap or the post-Friday close.

Expectations in backtest: with the 7-state regime filter, expect relatively few trades per week on M5 XAUUSD — most of the week is filtered out as CHOPPY or RANGE. The trades that do fire will be longer than a typical M5 trend bot because the 1R BE plus 2.5×ATR trail give runners room to breathe. The OnTester formula (net * pf) / (1 + dd) requires at least 30 trades to score non-zero, so short optimization windows will be heavily penalized. Use the tester with the full 90-day window or longer, with every tick or 1-minute OHLC, and the spread modeled as variable.

Pyramiding is the most interesting configurable behavior beyond the core signal. With InpPyramidEnabled true, the engine may add a second or third position in the same direction, but only when the existing position is already 1.5×ATR in profit. This is not a grid — each pyramid is a brand-new risk allocation that has to pass the regime gate, the HTF-EMA confirm, the RSI limit, the spread cap, and the daily/weekly loss limits. Most traders will leave it off; it is provided for users who want the option to scale into trends.