Pipsgrowth EX12016 MultiIndicatorConfluence

Five filters vote on every new M5 bar of XAUUSD, each with its own weight. The default scoring assigns Hull MA 10, SuperTrend 9, Heiken Ashi 8, RSI 6 and MACD 5, and the entry requires a combined confidence of 70% or higher on one direction. EX12016 is the weighted-vote member of the EX12 MultiIndicatorConfluence family — distinct from the 10-vote democracy of EX12005 (five adaptive MAs plus five oscillators on USDJPY M15) and from the ADX-bonus 5-vote of EX12007 (USDJPY M5/M30). This one reads five classical trend and momentum tools against the user's own weight table and only fires when those weights add up to 70% in one direction.

The Hull MA filter uses the standard Allan Hull construction: a 21-period WMA, a 10-period WMA, and the formula 2*WMA(period/2) − WMA(period). The function GetHullDirection(1, 21, 2) walks the last two completed bars and returns +1 only if Hull is rising on both, −1 if falling on both, and 0 otherwise. The InpHullRequired input is clamped to 1–3, so the EA can be tuned to demand a single rising bar (loosest), two consecutive (default), or three (strictest). The 21-period default is mid-range: short enough to keep up with XAUUSD's 5-minute rhythm, long enough to filter single-bar wicks.

The SuperTrend filter is the conventional two-close test, not the full SuperTrend indicator with its recursive state. CalculateSuperTrend(1, 22, 3.0) reads the most recent completed bar's ATR(22), computes the HL2 ± 3.0×ATR band, then asks: is the close above the lower band AND was the previous close also above the lower band? Two consecutive closes above the lower band vote bullish. The 3.0× multiplier is conservative for gold — most published SuperTrend presets use 2.0 or 2.5 — and 22-bar ATR covers roughly 1.8 hours of M5 data, smoothing out the session-open spikes that punctuate London and New York.

The Heiken Ashi filter does not read a built-in HA buffer; the EA reconstructs the bars itself. GetHADirection(2) walks the last two completed M5 bars, computes each bar's HA close as (O+H+L+C)/4 and the prior bar's HA open as (prev_open + prev_close)/2, and increments a bull or bear counter. Both bars must be the same color to vote. This is a slow, deliberate trend confirmation — the kind of filter that helps in trending London and New York gold sessions but cuts the EA's trade count during the Asian consolidation that often precedes the London open.

The RSI filter is a simple, non-crossing band check. It votes bullish when the 14-period RSI is above 50 but below the overbought level (default 70); bearish when RSI is below 50 but above the oversold level (default 30). The 50-line acts as a regime gate, and the OB/OS guards prevent voting into exhaustion. At weight 6, RSI is the second-lightest vote — the EA treats momentum as confirmation, not the primary signal.

The MACD filter is similarly reduced: it votes bullish only when the MACD main line is above its 9-period signal line AND the main line is also above zero. Symmetric for shorts. This means the EA does not fire long entries from negative-histogram crossovers, which is a deliberate choice — it requires both momentum (line > signal) and trend (line > 0) to agree. MACD's weight of 5 is the lightest, treating it as a tiebreaker rather than a driver.

The score is the sum of winning-side weights divided by the sum of all enabled weights, expressed as a percentage. A long signal needs at least 70% in the bull direction AND must beat the bear side; a short signal needs at least 70% in the bear direction. The default weight table (10 + 9 + 8 + 6 + 5 = 38 total) means a 70% threshold corresponds to 26.6 weight points, which is roughly Hull + SuperTrend + HA (27) without needing RSI or MACD, OR Hull + SuperTrend + RSI (25) without HA. The user can rebuild the voting system by toggling filters and reweighting — Hull, SuperTrend, and HA can each be turned off independently, as can RSI and MACD. The reweighting matters: dropping Hull to 5 and pushing SuperTrend to 12 makes the EA a SuperTrend-dominant system, not a Hull-dominant one.

Risk defaults to 1% of balance per trade with InpStopLoss = 500 points. The GetLotSize() function converts that risk percentage into a position size using tick value and tick size, snaps to the symbol's lot step, and caps the result at InpMaxLotSize = 10. The take-profit is fixed at 750 points — a 1:1.5 risk:reward ratio. The R:R is moderate rather than aggressive, and it matches the 70% confidence gate: with conviction around 70%, a 1.5R TP is the kind of math that survives realistic gold spreads and slippage without compressing the win rate into a coin-flip.

Pyramiding is on by default. CheckPyramidSafety(1) allows up to InpMaxPyramidTrades = 3 open positions on this magic, blocks adds if the existing position has secured less than 100 points in profit (calculated from the current SL, not from peak high water mark), and refuses opposite-direction adds. So a long can be added up to two more times, but only if the existing long's stop is already at least 100 points below its entry. That is the EA's interpretation of safe pyramiding — only add to a position that has already locked risk off the table, not to a position that is currently under water.

The trailing layer is ManageProfitLock(), a stepped ratchet that activates when a position is 300 points in profit. It then advances the stop-loss by 50 points at a time, always keeping 100 points of distance from the current price. This is a coarser, more conservative version of the typical ATR-based trail — it operates on fixed points, does not tighten as price moves (the 100-point distance is constant), and only fires when the position is already deep in profit. The user can disable the entire profit-lock layer by setting InpUseProfitLock = false, in which case the original SL stays in place until the TP hits or a manual close.

Time discipline is a server-time window, defaulting to 08:00–20:00, which covers the full London session and the first half of the New York session. Outside that window the EA simply does not open new trades; the existing positions are still managed by ManageProfitLock() and the spread/slippage guards. An optional InpCloseAtEnd = true will force-close all positions at the end-time, but the default leaves them alone — positions opened near 20:00 server can run into the next session if the user does not opt in to the close-at-end behavior.

The execution layer is a thin FOK wrapper. m_trade.SetTypeFilling(ORDER_FILLING_FOK) is set at init, and TryClose_EX12016() retries three times on REQUOTE, TIMEOUT, PRICE_OFF, or PRICE_CHANGED with a 200ms sleep between attempts. The modify path (TryModify_EX12016) retries three times on REQUOTE or TIMEOUT only, with a 100ms sleep — it does not retry on PRICE_CHANGED, because a moved price during a modify can leave the stop at the wrong level and the EA would rather reject the modify than misplace a stop.

The spread guard is hardcoded at 200 points (InpMaxSpread) — loose for XAUUSD by modern standards, so a low-spread broker matters. Slippage is set to 30 points (InpSlippage). The magic number is 22212016 and the trade comment is "Psgrowth.com Expert_12016" so the trades are easy to filter in post-trade analysis.

What you should expect in a backtest: a moderate trade count (the 70% confidence threshold cuts signals roughly in half compared to a single-indicator EA), most trades during London and New York overlap, and a healthy mix of wins and small losses with a 1.5R cap on winners. Because the EA is fixed-point (500pt SL, 750pt TP, 300pt trail trigger, 50pt trail step), the actual dollar risk scales with whatever the broker quotes — verify on your own symbol that "points" means what you think it means (typically 1 point = 1/10 pip on a 5-digit XAUUSD quote, so 500 points = 50 pips = $5.00 per 0.01 lot at gold's tick value). Run on a low-spread ECN broker with the server time zone set explicitly so the 08:00–20:00 window lines up with the sessions you actually want to trade.