Pipsgrowth EX10012 Momentum-Scalper

Pipsgrowth EX10012 is a single-position momentum-burst scalper for XAUUSD on the M5 timeframe, and its entire design rotates around one specific question: was the previous closed bar unusually large? The EA does not try to predict the next bar. Instead, the ComputeSignal() function inspects the last closed bar (shift=1) and checks whether its body — the absolute distance between open and close — exceeds the 14-period Average True Range multiplied by a configurable threshold (the InpBurstThreshold input, defaulting to 2.0). If yes, the EA classifies the bar as a momentum burst and records its direction. A confidence score is also produced: the ratio of body size to threshold feeds into a capped linear scale (strength - 1) * 50 + 50, with a ceiling of 100. The first burst bar to print at 1.5× ATR or higher lands at confidence 75, while a 2.0× bar reaches 100. This single fact — closed-bar body >= 2× ATR — is the only entry trigger. There is no cross, no crossover, no zone-retest logic layered on top.

Before that burst is allowed to become a trade, the EA routes it through a regime classifier (ClassifyRegime()) that divides the current market into seven labelled states: StrongTrend, WeakTrend, Range, Breakout, Compress, Expand, and Choppy. Classification is a two-step process. First, the 14-period ADX is checked: values at or above 28 mark a strong or expanding trend, and values between 18 and 28 mark a weak trend. Below 18 the market is treated as non-trending, and the decision falls to a 20-bar Bollinger Band width ratio: the current band width divided by the average width over the last 20 bars. Width below 75% of the average is labelled Compress (a squeeze, expected to precede a burst), width above 130% is labelled Range, and width in between is labelled Choppy (no trade). The ATR percentile over a 100-bar lookback contributes to differentiating StrongTrend (low percentile) from Expand (high percentile, ADX also >= 28). The classifier writes a bitfield — g_regimeFlags — with bits 1=trendOK, 2=rangeOK, 4=burstOK. The no-trade gate then refuses entries unless either bit 1 (trend) or bit 4 (burst) is set; Compress and Expand set the burst bit, StrongTrend and WeakTrend set the trend bit, and Choppy/Range are refused.

If the regime gate passes, three independent confirmations are applied in ConfirmDir(). First, the EA reads an H1 EMA(50) — the higher-timeframe trend filter — and refuses long entries when the bid sits below the EMA and short entries when the bid sits above it. This is the only HTF check; unlike dual-EMA cross-timeframe EAs in this family, EX10012 uses a single-EMA trend agreement. Second, the 14-period RSI on the M5 chart must agree: RSI(1) >= 50 for longs, <= 50 for shorts. Third, the MACD(12,26,9) histogram sign must match the trade direction; this is a soft filter — if the MACD buffer fails to load, the trade is allowed. All three checks can be disabled individually via the InpRequireRegime, InpRequireHTF, and InpRequireRSI toggles, which is useful for stress-testing how much of the EA's edge actually depends on each layer.

The stop-loss distance is computed from the same 14-period ATR that powers the burst detector: slDist = ATR(1) * InpSignalATRMult, defaulting to 2× ATR. The take-profit is set at slDist * InpTP_R (default 2.0R), so a 2× ATR stop pairs with a 4× ATR target. The minimum-reward-risk sanity check enforces InpTP_R >= 1.5 (the MIN_RR_RATIO define) — values below that print a skip message and refuse the entry. Position sizing is percentage-based by default: ComputeLots() reads the effective capital (capped by InpCapitalCapAmount if set, otherwise the live account equity), multiplies it by InpRiskPercent (default 0.5%), and divides by the per-lot loss at the chosen stop distance. The result is normalised to the symbol's volume step and clamped to min/max lot. A InpFixedLotMode toggle bypasses the percentage calculation and uses the symbol's minimum volume instead, which is the right setting when running on a sub-$1000 micro account or when validating behaviour without risking real capital.

Management of an open position is layered into four exit conditions, all running inside the per-tick ManagePositions() loop. The first is break-even: once the trade has moved 1R in profit (BE_TRIGGER_R), the stop is moved to entry plus 0.2R of locked profit (BE_LOCK_R). The second is a 50% partial close at +1R (PARTIAL_TP_R and PARTIAL_PCT=50); the partial is only sent if the remaining volume after the partial would still be at least 1.5× the symbol's minimum lot. The third is an ATR trailing stop: once the trade is up at least 0.8R (TRAIL_TRIGGER_R), the stop ratchets behind price at a distance of ATR(1) * 2.5 (the TRAIL_ATR_MULT define), and the new stop is only accepted if it actually improves on the existing one — there is no widening. The fourth exit is time-based: if a position has been open for 60 bars (five hours on M5, the MAX_BARS_IN_TRADE define), the EA force-closes it. A fifth escape hatch fires if the regime classifier labels the current state as Choppy — the EA closes the position immediately on the next tick, treating the trade as invalidated.

The risk stack around the entry is straightforward. NoTrade() evaluates seven distinct gates before allowing a signal to fire. Capital must be above the InpCapitalCapFloor (default $50). The current spread must be below InpMaxSpread (250 points — about 25 pips on a 5-digit gold quote, deliberately loose for an EA designed to fire on bursts). Daily realised loss must be below 3% of effective capital, weekly below 6% (InpDailyLossLimitPercent, InpWeeklyLossLimitPercent). The cooldown window (COOLDOWN_BARS=3, M5 = 15 minutes) only opens after InpCooldownLosses consecutive losses, detected by walking the last 10 deals from the deal stream inside UpdateLossStreak(). The deal history scan is limited to a 7-day window (TimeCurrent() - 7*86400). A position-count cap (InpMaxConcurrent, default 1) prevents pyramiding. The session gate blocks weekend trading entirely, refuses entries after 22:00 server time on Friday, and refuses entries before 02:00 on Monday.

There is no on-the-fly order retry beyond a single pass on TRADE_RETCODE_REQUOTE or TRADE_RETCODE_TIMEOUT. TryClose_EX10012, TryClosePartial_EX10012, and TryModify_EX10012 each loop three times on requote/timeout/price-off/price-changed responses with a 200ms sleep (100ms for modify), and they accept both TRADE_RETCODE_DONE and TRADE_RETCODE_DONE_PARTIAL as success. Filling mode is auto-selected from SYMBOL_FILLING_MODE in FOK → IOC → RETURN order. Slippage deviation is hardcoded to 20 points (MAX_SLIPPAGE_POINTS). The InpDryRun input defaults to true, so the EA will print the would-be trade — regime label, direction, confidence score — but never send an order; you must explicitly flip it to false to go live. The InpKillSwitch input, when set to true, halts every activity on the next tick and acts as a panic button for the operator.

Custom backtest grading is the last piece of the system. OnTester() returns (net * profit_factor) / (1 + drawdown_pct) provided the strategy executed at least 30 trades, has a positive profit factor, and a positive drawdown. Below 30 trades the criterion returns 0, which is the strategy-tester way of saying 'insufficient sample'. The OnTester formula is the same one used by other EAs in the Momentum-Scalper family, so the backtest rank of EX10012 is directly comparable to the other EX10 variants in MT5's optimisation pass — but EX10012's edge comes specifically from the burst-vs-regime interaction, not from a slow pullback or a state-dispatched signal, so the parameter ranges that produce meaningful wins will differ noticeably from its siblings.

The realistic operating picture: a single M5 chart on XAUUSD, on a low-spread ECN broker, with $100 minimum capital, sized at 0.5% per trade. Expect a low trade count per day — a 2× ATR burst on the M5 of gold is a real event, not a frequent one, and the regime gate will further prune trades during compressed or choppy phases. Most of the time the EA will be in a holding pattern, waiting for a bar that is clearly larger than its recent volatility. When one prints, the position opens and is actively managed: break-even ratchet, partial close, trail, time exit, and regime-change exit all run on every subsequent tick. The trade exits on its own terms, not on the operator's, and the input set is small enough to be walked through in a single MT5 strategy-tester optimisation pass without combinatorial explosion.