Pipsgrowth EX12074 MultiIndicatorConfluence

EX12074 is the only EA in the EX12 family that does not pin itself to a single symbol. It opens the chart, then spreads its attention across eight market instruments at once — XAUUSDm, XAGUSDm, USDJPYm, EURUSDm, GBPUSDm, AUDUSDm, EURJPYm and GBPJPYm — and picks the best opportunity per scan. The trading logic itself is built around a four-band Hull Moving Average consensus: at every new bar the EA computes Hull MA values for four different periods (periods are set per symbol preset — gold uses 14/21/34/55, silver and JPY pairs use 21/34/55/89, EUR/GBP pairs use 34/55/89/144, and any BTC/ETH symbol uses 10/16/26/42) using the formula WMA(2·WMA(price, period/2) − WMA(price, period), sqrt(period)). Each Hull's slope direction is checked against a five-point threshold to filter noise, and a direction is locked only when at least InpMinHullAgreement (default 2 of 4) point the same way AND the close sits on the correct side of the Hull. That agreement count is one of seven weighted pillars that combine into a single confidence score on a 100-point scale. The pillars are Direction (25 points — Hull agreement, higher-timeframe alignment, ADX strength, divergence-clean assumption), Timing (15 points — signal age under 20 bars, momentum consistency over 5 bars, optimal-session check), Asset Condition (10 points — ATR in 0.5x–1.5x of its 50-bar average, current spread within spreadMaxPips, regime classifier returns TRENDING), Filters (20 points — currently a flat 20/20 because InpRequireSRBreakout defaults to false), Entry Setup (15 points — 7 points assumed-confluence plus 8 from Hull-agreement count) and Exit Plan (15 points — valid SL/TP computable plus trailing plan armed). When CalculateConfidenceForSymbol() returns at least InpMinConfidenceScore (70 by default, 5–100 range), the symbol joins the candidate list. ScanAllSymbols() then sorts the list by confidence, takes the single best opportunity, and routes it to ExecuteTradeOnSymbol().

The market regime classifier behind the Asset pillar uses the symbol's ADX handle (iADX(sym, PERIOD_CURRENT, InpRegimeADXPeriod=14)) and ATR handle (per-symbol period from preset — gold 10, silver 12, JPY 14, EUR/GBP 14, BTC/ETH 8). Trending fires when ADX ≥ InpRegimeTrendADX (30 by default, with symbol-class overrides: gold 30, silver 28, JPY 25, EUR/GBP 25, BTC/ETH 35). Volatile fires when current ATR is at least InpRegimeVolatileATR (1.5× the 50-bar average). Quiet fires when ATR drops below InpRegimeRangeATR (0.5× the average). Anything else is RANGING. Although InpOnlyTrendingRegime exists in the input group, the scoring path no longer hard-blocks the non-trending regimes — the regime merely adds 0, 1 or 3 points toward the asset pillar.

The auto-calibration pipeline is the structural innovation. On OnInit, after symbol specs are read, CalibrateSymbol() runs once per enabled symbol. Step 3 calls MeasureSpreadStats() which samples the live spread for up to 5 seconds (or 100 samples at 50 ms intervals — whichever comes first) and records avg/min/max into SpreadStats. Step 4 calls CalculateATRStats() to capture the current ATR and a 50-period running average. Step 5 calls CalculateOptimalParameters() which derives symbol-specific distances in price units: SL distance = avgATR × InpStopLossATRMult (2.0) but floored at (stopLevel + avgSpread) × point × 1.5 so the broker's minimum stop is always respected, TP distance = SL × InpTakeProfitRR (2.0), breakeven distance = max(avgATR × InpBreakevenATRMult=0.5, SL × 0.3), trail start = max(avgATR × InpTrailStartATRMult=1.0, SL × 0.5), trail step = avgATR × InpTrailStepATRMult=0.5, and spreadMaxPoints = avgSpread × InpSpreadMaxMultiplier (2.5). Step 6 validates that SL ≥ stopLevel × point and that the breakeven distance is non-negative and within the SL range. When InpExportCalibration=true (the default), ExportCalibrationData() writes five CSV/log files to MQL5/Files: symbol specs, spread samples, ATR samples, optimal parameters, and a per-symbol validation log. The pipeline is one-shot at init; InpRecalibrateDaily is defined but the timer that would drive daily recalibration is commented out and all scanning runs from OnTick on new-bar events — the source comment explicitly states this was done so optimization and live trading use identical scan paths.

Dynamic risk scaling sits on top of the position sizer. The risk manager tracks consecutiveWins, consecutiveLosses, protectedBalance, peakBalance, currentDrawdown, safeModeActive and a daily P&L block. Base risk is InpStartRisk (25%), bumped by InpRiskIncrement (8%) per consecutive win, capped at InpMaxRisk (75%). When consecutive losses reach InpMaxConsecutiveLosses (3), the working risk is multiplied by InpLossReduction (0.5). When balance crosses 80% of InpTargetBalance (default $1,000), risk is multiplied by 0.5 even mid-trade. When balance actually hits the target, g_riskManager.safeModeActive flips true and risk drops to InpSafeModeRisk (3%) for the rest of the run. CalculateDynamicLotSize() then subtracts the protected capital (initial balance + 30% of any unrealized profit) before sizing the lot, so the tradable slice shrinks as the cushion builds. The result is fed into the standard risk formula: lotSize = riskAmount / (slDistance/tickSize × tickValue), then NormalizeLot() clamps to the broker's lotMin / lotMax / lotStep.

Capital-protection and correlation filtering are the final two layers. The CheckProtectionLimits() guard returns false (and the EA stops scanning for entries) when peak-to-current drawdown ≥ InpMaxDrawdown (25%) or when intraday loss ≥ InpMaxDailyLoss (15% of the day's starting balance). CheckSymbolProtections() short-circuits a single symbol after 3 consecutive losses on that symbol or when the live spread exceeds the calibrated spreadMaxPoints. IsCorrelatedWithOpenPositions() walks all open positions whose magic matches InpMagicNumber=22212074, looks up the pair in g_correlationMatrix, and blocks the new opportunity when the absolute correlation exceeds InpMaxCorrelation (0.70). The correlation matrix is built once at init by BuildCorrelationMatrix() using EstimateCorrelation() — a hardcoded lookup: any two symbols containing "EUR" return 0.8; EURUSD vs GBPUSD return 0.85; any two USD-pairs return −0.5; any two JPY pairs return 0.7; XAU vs XAG return 0.6; everything else returns 0.3. This is a deliberate approximation, not a rolling correlation computation, and is noted in the source.

Exit management runs every tick on every position. ProcessAllExits() walks the EA's positions in reverse, then for each one calls MoveToBreakeven(), PartialClosePosition() and TrailStop() in that order. MoveToBreakeven() uses the calibrated beDistancePrice — once price moves that far in profit and the current SL is still on the wrong side of entry, SL is moved to entry+point for buys or entry−point for sells via TryModify_EX12074(). PartialClosePosition() checks a per-ticket global variable (PartialClosed_<ticket>) so it fires only once, then closes InpPartialClosePercent (50%) at the broker minimum lot when priceDiff ≥ slDiff (i.e. profit has reached 1:1 R:R). TrailStop() activates once price has moved trailStartPrice in profit, then ratchets the SL to currentPrice ± trailStepPrice, but only when the new SL is meaningfully tighter than the current one (newSL > sl + trailStep × 0.5, or current SL is zero). All three exit modifications route through the TryModify_EX12074() retry wrapper, which retries on REQUOTE/TIMEOUT/PRICE_OFF/PRICE_CHANGED with 100 ms backoff.

What the EA does NOT do: it does not run a timer (the OnTimer function is a stub that returns immediately), it does not apply a global hard equity-stop inside the trade functions (the only stops are the per-position SL/TP plus the per-symbol/drawdown gates in CheckProtectionLimits), it does not grid or martingale (no order layering, no lot multiplier on loss), and it does not persist a running correlation coefficient — the matrix is static after init. The dashboard (InpEnableDashboard=true by default) writes a banner via Comment() showing balance, equity, target, peak, drawdown, mode (AGGRESSIVE/SAFE), win streak, win rate, top 5 symbol stats, and last-scan time, and is suppressed during MQL5 optimization passes so it does not pollute the optimizer log.