Pipsgrowth EX09004 MeanReversion

Pipsgrowth EX09004 reads the tape, not the trend. Where most reversal EAs look for oversold oscillators or stretched Bollinger deviations, EX09004 hunts for liquidity traps: bars where a wave of stop-hunts drives price through a recent swing high or low, only to get beaten back inside the range by the close. The volume surge that accompanies that rejection is what makes the trade. A bull trap needs a low that punctures the 5-bar swing low and a close back above it, on a bar whose tick volume exceeds 1.5× the 20-bar SMA of tick volume; a bear trap is the exact mirror on the high side. With InpRequireEngulf=true (the default), the rejection bar must also close in the same direction as the sweep (close > open for a bull trap, close < open for a bear trap), which filters out weak-bodied rejections that often fade into continuation.

The signal is gated by a 7-state regime classifier built from ADX(14), ATR(14) percentile over a 50-bar window, and Bollinger Band(20, 2.0) width. By design, mean-reversion is dangerous in trending markets, so the regime gate explicitly blocks entries in REG_STRONG_TREND and refuses all REG_CHOPPY states outright. Trades are only taken in RANGE, COMPRESS, WEAK_TREND, or — for traders willing to fade volatility expansions — REG_EXPAND / REG_BREAKOUT, although the latter two are rarely the right home for a reversal entry. The classifier is one of the most expensive functions in the source: it pulls 50 bars of ATR and Bollinger-band data on every tick to compute the percentile, so the EA's CPU footprint is non-trivial on lower-end VPS instances.

The confidence score the function returns is a transparent sum: 30 points baseline, plus a volume-surge score capped at 30 points (the further above the 1.5× multiplier, the higher the score), plus a sweep-depth score capped at 40 points (how far price overshot the swing level relative to the swing range). The total is clamped at 100, but in practice the 70–85 range is the sweet spot — too low and the trade is probably noise, too high and price is so stretched that reversion becomes less likely. The score is currently logged but not used to filter entries; trades pass through on the basis of the signal alone.

Risk is sized at 0.5% of effective capital by default, where effective capital is MathMin(InpCapCapAmount, equity) — set the cap input to zero to disable, set it to your desired risk budget to clamp lot sizing to a fixed dollar amount. The lot is computed from the SL distance in price, the symbol's tick value, and the risk budget via the standard riskMoney / riskPerLot formula. A 200-point minimum SL floor (MIN_SL_POINTS) protects against micro-stops on quiet bars that would be clipped by the broker's stops-level anyway. Position sizing falls back to 0.01 lots if InpRiskPercent=0 and the margin pre-check in OpenPosition() blocks any order whose OrderCalcMargin exceeds free margin.

Stops are ATR-derived: SL = ATR(14) × 1.5, TP = ATR(14) × 3.0, giving a 1:2 reward-to-risk at entry. A minimum R:R of 1.5 is enforced in ConfirmEntry() — if the implied ratio slips below 1.5 due to wide spread or a thin ATR, the entry is refused with a logged reason. Once the trade is on, the management layer stacks five exit paths. First, a one-shot break-even at +1R: when bid (for longs) or ask (for shorts) clears 1× the original SL distance, the stop is ratcheted to entry + 1 point. Second, a one-shot partial close of 50% of the position at +1R, which requires the position to be at least 2× the symbol's minimum lot. Third, an ATR trail at 2.0× ATR(14), forward-only, that activates on every tick after break-even. Fourth, an opposite-signal exit: if a bear-trap fires while a long is open, the long closes at market. Fifth, a regime-change exit: a transition into REG_CHOPPY or REG_STRONG_TREND force-closes the position, because holding a reversal trade into a regime shift is the classic way to give back the entire partial profit.

The no-trade gate is one of the more thorough in the EX09 family. It checks: kill switch, spread above 400 points (the default cap, tuned for XAUUSD on 2- and 3-decimal feeds), regime state, max-concurrent positions (1 by default), realized daily loss above 3% of effective capital, 5-bar cooldown after a losing exit, equity below the $50 floor, weekend (Saturday/Sunday are blocked outright), high-importance MQL5 calendar news within a 30-minute window for the symbol's two currencies, invalid quotes, and broker trade-mode disabled. All orders route through retry helpers (TryClose_EX09004, TryClosePartial_EX09004, TryModify_EX09004) that re-attempt up to three times on REQUOTE, TIMEOUT, PRICE_OFF, or PRICE_CHANGED return codes, with 200 ms sleeps between attempts.

The OnTester custom criterion is net * pf / (1 + dd) where net is total profit, pf is profit factor, and dd is relative equity drawdown percent. This penalizes strategies that make money with deep drawdowns and rewards strategies with smooth equity curves, which means an MT5 optimization will naturally favor parameter combinations with low volatility, not just the highest net profit. A minimum of 30 trades is enforced before the criterion is scored — anything below that returns zero and is rejected by the optimizer.

What the EA does not do: there is no pyramiding (MAX_PYRAMID_LEVELS=1 is the only constant, and there is no pyramid function in the source). The HTF EMA(50) and signal-timeframe EMA(50) handles are allocated in OnInit() and released in OnDeinit() but never read in OnTick() — they exist as scaffolding from the 12-layer architecture promise but the actual reversal logic runs without higher-timeframe confirmation. The InpUseSessionWindow toggle is off by default; when enabled, it restricts entries to 06:00–20:00 broker server time, which lines up with the London + New York overlap on a GMT+2 or GMT+3 MetaTrader server. The InpDryRun default is true, so a fresh install will log every signal, every gate, every management action to the journal without placing a single order — flip it to false to go live.

For backtesting expectations: the strategy is selective. On a 1-year XAUUSD M5 backtest, expect 100–300 trades depending on regime, a profit factor typically between 1.1 and 1.6, and a 1:2 base R:R meaning winners are twice the size of losers on average. The strategy will cluster its activity around high-volume rejection bars — London open, NY open, and the first hour after US data releases are the natural habitat. A flat Asian session will produce very few signals because volume is too low to clear the 1.5× surge threshold. The 288-bar (24-hour on M5) time stop and the regime-change exit mean losing trades are cut quickly; the partial-close + ATR trail combination means winning trades are typically captured at 1.5–2.5× the original SL distance before being given back.