Pipsgrowth EX18053 TrendFollow

EX18053 — codename Hull2EA — is a 4-line inline Hull moving-average voter that only opens a position when at least three of the four slopes agree on direction, and the remaining lines have not registered a vote in the opposite direction. The four Hull periods are 20, 34, 55, and 200 bars, all calculated directly on close[] inside the EA using a Linear Weighted Moving Average primitive rather than the iCustom/iMA pipeline. The LWMA primitive is straight sum-of-weighted-prices / sum-of-weights with descending weights (most recent bar gets the heaviest weight), and the Hull itself combines a fast LWMA at period/2 with a slow LWMA at the full period, then smooths the difference with a sqrt(period) weighted average. The result is a low-lag, low-noise trend line that flips between an up-slope and a down-slope without the cross-over delay a plain EMA would impose. The HullColor helper returns 0 when the current Hull value is greater than the previous one, 1 when smaller, and 2 when they tie — and the EA only counts votes 0 or 1, so a flat line contributes nothing and can never break a tie.

The consensus rule is therefore stricter than a simple majority: with the default InpMinAgree=3, longs require at least three of the four Hulls to be sloping up AND none sloping down (shortVotes must be 0). The same is true in reverse for shorts. That asymmetric rule means a single contradictory line in the opposite direction blocks the signal entirely, even if the other three agree — a design choice that filters out the 3-up-1-down whip-saw patterns that simpler vote schemes allow. A confidence figure is also produced, anchored at 50 and then nudged up by the magnitude of the fast Hull slope relative to ATR (50 + slope/ATR * 200, capped at 100), giving a rough sense of how decisively the fastest line is moving.

Two confirmation gates sit in front of entry. ConfirmADXStrong() requires the most recent closed-bar ADX(14) value to be at least 18 — the same threshold that defines WEAK_TREND in the regime classifier — so the EA will not open on a flat, range-bound tape even if all four Hulls happen to be sloped. ConfirmMinRR() blocks entries if TP/SL would be below 1.0, which the 3.2/1.8 default ATR multipliers comfortably clear. Optionally, HTFHullBias() runs the same Hull(35) calculation on the higher-timeframe close[] (default H1) and demands that it slope in the same direction as the entry — a 5-minute long signal against a 1-hour Hull rolling over is rejected before any sizing is computed.

The regime classifier runs every tick and feeds the no-trade gate. It is a seven-state machine driven by ADX(14), a 50-bar ATR percentile, and a 50-bar Bollinger-band width percentile. STRONG_TREND fires when ADX is at or above 25 and the current ATR sits in the top quartile of its 50-bar distribution. WEAK_TREND requires only ADX >= 18. COMPRESS is the band-width-percentile floor (<=20), EXPAND the ceiling (>=80). BREAKOUT requires high ATR (top quartile) without the ADX strength to qualify for STRONG_TREND — a volatility impulse the trend indicators have not caught up to. RANGE is a 40-70 band-width window with weak ADX. Everything else becomes CHOPPY, which is the only regime that outright blocks new entries. COMPRESS, EXPAND, BREAKOUT and RANGE all allow entries; STRONG and WEAK are the productive states.

The no-trade stack that follows the regime gate is a sequence of early returns, each logged with a reason. SpreadOK() rejects entries when the current spread exceeds 0.45 * ATR(14) — an ATR-relative spread cap that automatically relaxes in volatile conditions and tightens in quiet ones. SessionOK() simply blocks Saturday and Sunday entirely and refuses new entries after 21:00 on Friday, so a position is never opened into a weekend gap with no operator watching. DailyLossOK() sums the EA's own realized PnL for the day (filtering HistoryDeals by magic and symbol) and blocks if the cumulative loss has reached InpDailyLossPercent (default 3.0%) of effective capital. IsCapitalBelowFloor() blocks when effective capital has fallen below InpCapitalCapFloor (default $50), which is a useful guard against a blown account still firing orders. Max concurrent positions is capped at InpMaxConcurrent (default 3). Margin is verified up-front via OrderCalcMargin before any position is opened, so a rejected margin check never reaches the trade layer. Finally, a duplicate-entry guard tracks the last entry bar and direction and refuses to re-open the same direction on the same signal bar.

Sizing is fixed-fractional on the effective capital figure produced by ComputeEffectiveCapital(), which takes the minimum of account equity and InpCapitalCapAmount when the cap is non-zero, and uses pure equity otherwise (cap=0 disables it). The default 0.5% risk is converted to a stop-distance-aware lot count by dividing the dollar risk by the per-lot stop-distance loss calculated from the symbol's tick value and tick size. If InpRiskPercent is set to 0 the EA falls back to InpFixedLots (0.10), and NormalizeLots rounds the result to the broker's volume step and clamps to the min/max lot range. The whole sizing chain is therefore broker-portable: the same 0.5% risk produces different lot counts on a 5-digit gold broker and a 4-digit FX broker, but the dollar risk is identical.

Once a position is open, the ManageOpenPositions() function runs on every tick. Three management actions can trigger. The break-even one-shot fires when price has moved in the trade's favour by ATR * 1.0 (= 1R) and ratchets the stop to open + 1 point; the gate is sl < newSL so the BE only moves forward, never backward. The partial-TP at TP1 fires once per position at the same 1R profit trigger and closes half the position (PARTIAL_TP_FRACTION = 0.5), with a vol > g_volMin guard so a sub-min partial does not get rejected by the broker. The trailing stop activates when price has moved 2 * trailDist = 3R in profit, and ratchets the stop to bid/ask - ATR*1.5 from there on. A 240-bar time stop (MAX_BARS_IN_TRADE, 20 hours on M5) forcibly closes any position that has not produced either a TP hit or a stop hit by then.

Exit conditions come in two flavours beyond the SL/TP and the trailing/partial machinery. CheckRegimeExit() fires on every new bar; if the regime has flipped to CHOPPY, all EA positions are closed. If InpCloseOnReverse is true (the default) and the consensus signal has flipped to the opposite direction of the held position, that direction's positions are also closed. A separate time-exit path inside ManageOpenPositions() closes positions that exceed MAX_BARS_IN_TRADE bars in age. There is no news-window filter, no Friday-cutoff close (only a Friday-cutoff new-entry block), and no end-of-day force-close.

Pyramiding is implemented but disabled by default (InpEnablePyramid = false). When enabled, CanPyramid() allows up to PYRAMID_MAX_LEGS = 3 positions in the same direction, provided each leg is opened at least 0.5 * ATR beyond the previous leg's open price and the previous leg is currently in profit with a stop already moved to breakeven or better. Combined with the 1R BE rule, the practical effect is that pyramiding stacks winners that have survived the initial 1R pullback, rather than adding into losers.

The capital allocation cap is a two-input subsystem. When InpCapitalCapAmount is greater than zero, effective capital is the lesser of the configured cap and current equity, and IsCapitalBelowFloor() refuses new entries below InpCapitalCapFloor (default $50). This is the closest thing the EA has to a hard account-level stop and is most useful for traders who run multiple instances on the same account and want each instance to pretend it has its own dedicated sub-account. With the cap at the default 0.0, the EA uses the full account equity for sizing and the floor check never triggers.

Two operator toggles sit at the bottom of the input panel. InpDryRun defaults to true — every intended order is logged to the journal (DRYRUN BUY/SELL lots=... price=... sl=... tp=...) but no order is actually sent to the trade server. This is a deliberate safety default: the EA ships in a paper-trade state and the operator has to deliberately flip DryRun to false to go live. InpKillSwitch, when true, short-circuits the OnTick flow before any signal processing and just refreshes the dashboard — a panic-button that can be toggled without recompiling.

In a backtest, the strategy behaves like a typical 4-line Hull-MA crossover voter with conservative risk management. The custom OnTester criterion multiplies net profit by profit factor and divides by (1 + balance drawdown), and disqualifies any test with fewer than 30 trades — so a curve-fit over a handful of trades cannot win the optimization pass. Live, the entry frequency is bounded by the regime gate and the 3-of-4 vote rule; a choppy market produces very few signals, and a strong trend can produce one or two per day on M5 XAUUSD. The combination of the 1.78:1 R:R (3.2/1.8) and the regime filter is what makes the system selective: it prefers not to trade rather than to trade badly.