Pipsgrowth EX16062 Trend

Adaptive confluence engine built around a five-signal weighted voting system, where the Adaptive Moving Average (AMA) carries the heaviest weight and a self-tuning sensitivity knob rescales the confirmation threshold in response to recent win/loss streaks. EX16062 is not a single-indicator trend follower — it is a layered scoring model that adapts the entry bar based on what the market has just been doing.

The five signal sources and their weights in GetMarketSignal() are fixed at the source level: AMA = 0.4, RSI = 0.2, MACD = 0.2, Stochastic = 0.1, and a price-action trend filter (SMA(5) vs SMA(20)) = 0.1. Each component casts a discrete -1, 0, or +1 vote, and the EA enters long when the weighted sum crosses above +0.5 * adaptiveSensitivity and short when it falls below -0.5 * adaptiveSensitivity. That sensitivity is not a static input — it is a runtime variable (adaptiveSensitivity, default 0.5, clamped to the [0.3, 1.0] band) that drifts up by 0.1 after three consecutive wins and down by 0.1 after two consecutive losses. The practical effect: the EA tightens its confirmation requirements when it is losing and loosens them when it is winning. A long stretch of profitable trades will let a barely-confirmed AMA crossover through, while a losing stretch demands three or four components to agree before any new position is sized.

The AMA itself runs with AMA_Period=14, AMA_FastEMA=2, AMA_SlowEMA=30, and the entry fires when price crosses the AMA line in either direction (current bid above AMA[0] while previous close was below AMA[1] for longs, and the mirror for shorts). The RSI layer is not a generic 30/70 — it is volatility-adjusted. The EA computes volatilityRatio = currentATR / averageATR(3) and shifts the overbought threshold up by (volatilityRatio - 1) * 10 * adaptiveSensitivity and the oversold threshold down by the same amount. In a high-ATR regime, RSI must run more extreme to register a buy or sell vote; in a quiet session, ordinary 30/70 excursions are sufficient. MACD is a standard 12/26/9 line-vs-signal crossover; Stochastic is a fixed 20/80 zone reading on the %K line with no smoothing for an overbought/oversold vote. The trend filter uses two simple moving averages that are computed on the fly inside the EA — period 5 and period 20 — and contribute a directional vote, not an entry trigger.

Position management is a strict one-position-per-symbol model. The OnTick block reads the current market signal, and if the signal conflicts with the open position, EX16062 first deletes any pending orders via ClosePendingOrders(), then closes the opposing position with CloseAllPositions() before opening the new direction. There is no grid, no martingale, and no averaging — the EA does not tolerate two positions in flight at once, and it does not build a basket. On a signal that matches the open direction, the EA does nothing new (the existing trade continues to manage itself).

The stop loss and take profit are ATR-driven, not fixed-pip. SLMultiplier=1.0 and TPMultiplier=2.0 mean SL = ATR(14) * 1.0 and TP = ATR(14) * 2.0, giving a fixed 1:2 reward-to-risk at the moment of entry regardless of where ATR is sitting. A 50-point ATR pair gets a 50/100 stop/target; a 200-point ATR pair gets a 200/400 stop/target. Lot sizing is risk-based: CalculateLotSize() divides RiskPercent * balance by stopLossPoints * tickValue, then rounds down to the broker's SYMBOL_VOLUME_STEP and clamps inside SYMBOL_VOLUME_MIN / SYMBOL_VOLUME_MAX. With RiskPercent=1.0 on a $100 account, a 100-point SL would size to roughly 0.01 lot on a $10/pip pair and 0.10 lot on a $1/pip pair.

The trailing stop is mechanical and runs from ManagePositions() on every tick. After the position has moved TrailingStart=15 points in profit, the EA shifts the stop by TrailingStep=5 points behind the new bid (for longs) or above the new ask (for shorts). The check is newSL > currentSL for buys and newSL < currentSL || currentSL == 0 for sells, so the ratchet is unidirectional and will never move the stop against the position. The trailing stop only modifies the SL; it does not touch TP. A position can still run to its original 2x ATR target while the stop is being ratcheted behind it.

The risk stack is conservative for a 25-input EA. IsSafeToTrade() runs four gates before any signal evaluation: equity must stay above MinEquityPercent=95% of the starting balance (otherwise the EA calls CloseAllPositions() and prints a hard stop), realized daily P&L must not have hit MaxDailyLossPercent=3%, trades opened today must be below MaxTradesPerDay=5, and the active session must contain the current server hour. If any gate fails, the EA returns false and the tick is a no-op. There is no kill switch or recovery — once the daily loss trips or the equity floor trips, the EA flattens and sits out for the rest of the day (the counters reset at the next date rollover in UpdateDailyCounters()).

The session filter is a simple hour-of-day window. SessionStartHour=2 and SessionEndHour=16 define the tradeable band (server time), which the source comments as "London session start" to "New York session end." IsInTradingSession() correctly handles a window that crosses midnight via an OR branch, although the defaults do not require that branch. The news filter is a simplification — IsNewsTime() returns true when the current time is within 30 minutes of 08:00 or 13:00 server time, which the source labels as the London and New York opens. It is not connected to an economic calendar feed; high-impact data releases outside those two fixed windows are not filtered, and the windows themselves are timezone-sensitive to the broker's server clock.

The CloseAllPositions(), ClosePendingOrders(), and trailing updates all build their own MqlTradeRequest / MqlTradeResult structures inline and check result.retcode == TRADE_RETCODE_DONE before logging success or the human-readable error string from GetTradeErrorDescription(). Three retry helpers (TryClose_EX16062, TryClosePartial_EX16062, TryModify_EX16062) are also defined for wrapper-based operations with 200ms backoff on requotes, timeouts, and stale prices — but they are not called from OnTick() or from the close / modify / partial-close paths above. They are dormant wrappers and would only matter if a future code path invoked them; for now the only paths that touch the broker are the inline OrderSend() calls in OpenBuyOrder(), OpenSellOrder(), CloseAllPositions(), ClosePendingOrders(), and TrailingStop().

The OnTester() hook is not implemented in this source — no custom fitness function is present. The strategy is meant to be optimized via MetaTrader's standard parameter sweep over the indicator periods and risk percentages, not a custom objective. The fill policy is the broker's default, since the EA never sets request.type_filling explicitly. On netting accounts the EA may issue market orders with no type_filling override, which works on ECN and classic hedging servers but can fail on some exchange-style symbols that require ORDER_FILLING_FOK or ORDER_FILLING_IOC.

For live deployment, the EA is hard-coded to one position at a time on _Symbol, so a $100 minimum-deposit account with RiskPercent=1.0 is workable, although the source comment "max 2% for $100 account" suggests the developer expects a 1-2% risk band on the smallest accounts. The trailing stop and ATR-driven SL scale with volatility, so on a quiet range-bound XAUUSD session the stops will be tight, and during a news-driven spike the stops will widen on the next entry. The adaptive sensitivity knob will also ratchet the threshold down after a 2-loss streak, which means losing days will produce more signals, not fewer — a deliberate bias toward "try again sooner." The two retry helper functions are present but unused, so any session-level slippage handling beyond the inline OrderSend() checks has to be added externally.