Pipsgrowth EX11002 MultiFilterScalper

EX11002 MultiFilterScalper is built around a single architectural decision that sets it apart from typical indicator confluence EAs: instead of requiring every filter to agree (an AND-gate) or triggering on any single filter (an OR-gate), it tallies votes from each enabled filter and only fires when a majority of the active filters point in the same direction. With the default configuration — EMA trend, ATR range gate, Heiken Ashi, and SuperTrend all enabled, Hull MA disabled — that means the EA waits for three of the three directional filters (EMA, HA, SuperTrend) to all agree on a buy or sell before it commits capital. This majority-vote structure trades a small amount of trade frequency for a meaningful reduction in false triggers, because the EA is structurally incapable of entering on a one-filter flicker.

Each filter contributes a single integer vote on every closed bar (shift=1 to prevent repainting). The EMA filter reads the close of the previous bar against a 50-period exponential moving average and votes +1 if price is above the EMA, -1 if below. The Heiken Ashi filter rebuilds the synthetic HA candles inline and requires two consecutive bullish HA closes for a buy vote, two consecutive bearish for a sell vote, with the count controlled by InpHaConsecutive. The SuperTrend filter computes the classic upper/lower band — high-low midpoint plus or minus three ATR multiples — over a 10-bar ATR window and votes +1 if the current and previous closes are both above the lower band, -1 if both are below the upper band. The Hull MA filter, disabled by default and activated by setting InpUseHullFilter = true, builds the standard Hull construction (2×WMA(half-period) minus WMA(full-period)) and votes on the direction of the Hull value from one bar to the next.

The ATR filter is structurally different from the other three. It does not cast a directional vote — instead it acts as a range gate. The EA reads the current ATR(14) in points and refuses to trade when ATR is below InpMinAtr (default 0.5 points, meaning the market is too quiet) or above InpMaxAtr (default 50.0 points, meaning volatility has spiked into a regime the EA does not want to participate in). Setting InpMaxAtr = 0 disables the upper bound. This is the EA's primary environment filter, and it does not contribute to the vote count; it simply returns from OnTick before the voting logic runs.

The required vote threshold scales with the number of enabled directional filters. With Hull off, the EA counts EMA, HA, and SuperTrend (three filters) and needs at least total/2 + 1 = two of the three to agree. With Hull on, that becomes three of four. Crucially, the EA also checks that the majority direction strictly exceeds the minority — so two buy votes and one sell vote on a three-filter configuration will not fire on a buy. This is the difference between "majority agrees" and "majority is decisive."

When a signal does fire, the EA computes the stop and target as multiples of the current ATR. InpStopLossAtr = 1.5 and InpTakeProfitAtr = 2.0 give a stop-to-target ratio of 1.5:2.0 (R-multiple of 1.33), with a hard floor of 50 points on the stop distance to keep the broker's SYMBOL_TRADE_STOPS_LEVEL constraint from rejecting the order. GetLotSize() then sizes the position to risk InpRiskPercent = 1.0% of effective capital against that stop distance, with InpFixedLot = 0.01 as the fallback when risk-percent is set to zero, capped at InpMaxLotSize = 10.0. The capital base passed to the lot calculation is GetEffectiveCapital(), which by default returns the full account equity but can be clamped to InpCapitalCapAmount when that input is non-zero — a deliberate opt-in for traders who want to ring-fence a portion of their balance to this EA only.

The execution path is straightforward: the EA requests a buy at the current Ask or a sell at the current Bid with the ATR-derived SL and TP, checks that the order's required margin does not exceed the free margin, then attempts the order in a three-iteration retry loop with 150ms sleeps on REQUOTE, TIMEOUT, and PRICE_OFF returns. The CTrade instance is configured to use the broker's preferred filling mode selected in FOK > IOC > RETURN order from the SYMBOL_FILLING_MODE mask at init time. The slippage tolerance is 30 points (InpSlippage).

Position management runs on every tick, not just on new-bar logic. ManageProfitLock() walks every open position for this EA's magic number and, once profit exceeds InpLockTrigger = 300 points, sets a new stop at the current price minus InpLockDistance = 100 points — but only advances the stop in discrete 50-point steps via InpLockStep, never backwards, which prevents the stop from loosening during a pullback. The same function handles both long and short positions. This step-ratchet style is intentionally simpler than a continuous trail: it produces fewer order modifications, which means fewer re-quote risks and less broker load.

Pyramiding is on by default (InpUsePyramid = true) and supports up to InpMaxPyramidTrades = 3 open positions. The CheckPyramidSafety() function blocks any add-on when the new signal would reverse the existing direction (a buy while a sell is open, or vice versa), and crucially it measures secured profit — the distance between the current stop and the entry price — not floating profit. A new position is only added when the existing position's stop has already been advanced at least InpPyramidMinProfit = 100 points into profit, ensuring every pyramid addition is into a trade that has already proven itself. With InpUsePyramid = false, the EA falls back to a strict one-position-at-a-time mode.

The hardening stack is unusually comprehensive for a single-MA-filter scalper. IsSafeToTrade() returns false when any of the following conditions hold: equity is at or above the capital cap, equity has dropped below the InpCapitalCapFloor = $50 floor, equity has fallen below 95% of the initial balance, daily realized loss exceeds 3% of effective capital, weekly realized loss exceeds 6%, the consecutive-loss counter has hit 3 and the cooldown timer is still active, the daily trade count has reached 50, the market is closed, the current GMT-derived time is outside the configured session windows, or the optional news filter is on and the time is within 15 minutes of a London or New York open. Realized PnL is fed into these counters through OnTradeTransaction, which walks the deal history on every DEAL_TRANSACTION_DEAL_ADD event and updates g_realizedToday, g_realizedWeek, g_tradesToday, and g_consecLosses from the actual fill records, not from EA bookkeeping. After three consecutive losses, a 30-minute cooldown (InpCooldownMinutes) blocks all new entries regardless of signal strength.

The session filter is GMT-based, not server-time-based. DetectGMTOffset() walks back up to 100 H1 bars at init looking for the weekend gap (the jump from Friday 22:00 to Sunday 22:00 or Sunday 23:00) and computes the broker's offset from GMT, or accepts the manual override via InpServerGMTOffset. The active session is then computed in real GMT: London 7-16, New York 12-21, and Asian 0-7. With InpAvoidAsia = true (the default), the EA refuses to trade during pure Asian hours, but it will still trade the London and New York overlap, which is the session XAUUSD scalpers typically want.

Time-based exits are not built into this EA in the conventional sense — there is no InpMaxHoldBars input. The mechanism that closes positions is the profit-lock ratchet (which guarantees that any trade that moves 300+ points in profit will eventually close on a pullback) combined with the optional InpCloseAtEnd = true flag, which closes all positions for this magic number when the server time crosses the end of the configured trading window. The default time window is 08:00-20:00 server time, which roughly brackets the London and New York sessions depending on broker offset.

The OnTester() fitness function for strategy optimizer runs is (net × pf) / (1 + dd) with a minimum of 30 trades required for a non-zero result, and it short-circuits to zero when the trade count is below that floor. The divisor protects against degenerate results on small samples. This is the same fitness family used across the EX11 family.

In backtesting and live deployment, this EA is a slow-twitch, high-conviction XAUUSD M5 scalper. The vote structure means it will not fire on every 5-minute bar, even in clean trends — it will fire only when at least two of EMA, HA, and SuperTrend agree on direction, with ATR confirming the environment is neither dead nor chaotic. The profit-lock ratchet means that winning trades routinely give back the last 100 points before exiting, which is the price of the simple-step trailing design. The hardening stack is the main reason to consider this EA on a live account: the daily and weekly loss caps, the cooldown, and the equity floor are all running automatically and cannot be overridden by signal strength.