Pipsgrowth EX10008 Momentum-Scalper

Pipsgrowth EX10008 Momentum-Scalper is a regime-dispatched XAUUSD scalper that classifies each closed bar into one of five market states — TREND_UP, TREND_DOWN, RANGE, VOLATILE, or QUIET — and then routes to a different signal logic depending on which state it finds. The classification is computed inside ClassifyMarketState() from a fast EMA(9), a slow EMA(21), the current ATR(14), and the Bollinger Band middle line: when the ATR-to-midline ratio is below 0.1 the bar is QUIET (no trading at all), when it is above 0.5 the bar is VOLATILE (a hard regime that requires triple-confirmation), and when |FastEMA − SlowEMA| exceeds half of ATR the state is one of the two trend states. Everything else is RANGE. This dispatching is the architectural centre of the EA: rather than applying a single signal formula to every bar, the code switches on g_marketState and uses a tailored entry condition for each branch.

In the TREND_UP branch the EA looks for a fast EMA crossing above the slow EMA on the closed bar, or alternatively an oversold RSI(7) reading (below 25) combined with a close above the slow EMA. The mirror condition fires in TREND_DOWN: fast EMA crossing below slow, or RSI(7) above 75 with a close below the slow EMA. The RANGE branch is a mean-reversion read: a touch of the lower Bollinger Band (with a 0.1% tolerance) combined with either an oversold RSI or a bullish RSI divergence fires a buy, and the upper-band touch fires a sell. VOLATILE is the strictest branch — it requires a three-factor AND: the appropriate EMA cross, the matching RSI extreme, and the matching Bollinger touch. QUIET is silent; the default case returns no signal. The confidence score attached to each signal is also state-specific: trend signals use min(100, |EMA-fast − EMA-slow| / ATR × 50 + 25), range signals use (25 − RSI) × 3 + 40, and the volatile branch sets a flat 90.

Before any state-dispatched signal reaches the entry path it must clear two global filters. IsVolumeSpike() requires the previous bar's tick volume to exceed 1.5× the average of the prior five bars. HasSufficientMomentum() requires the absolute FastEMA delta between the last two closed bars to be at least 50 points. Both toggles can be switched off via the InpUseVolumeFilter and InpUseMomentumFilter inputs, but they default to true because they are the principal protection against false breakouts in fast gold moves.

A second, redundant confirmation sits between the signal and the order: RegimeOK() reads ADX(14) and rejects entries when the main ADX line is below 18. It also reads the last 52 ATR values to compute an ATR-percentile against its own moving average, and rejects trades when the percentile is below 15 (dead tape) or above 95 (news spike). A third sub-check reads the last 10 Bollinger-width readings and rejects when the band is in a tight squeeze below 10% of its own moving average. The optional InpHTFConfirm input then layers an M15 EMA(50) / EMA(200) agreement check on top — when the M15 fast EMA is below the slow EMA, no buy fires, and vice versa. This HTF layer is the most consequential single input in the EA: it does not change what is computed, it only gates whether the result is allowed to send an order.

Position sizing is risk-percentage based on InpRiskPercent (default 0.7% of effective capital) with a InpFixedLot fallback of 0.01 lots when risk calculation is invalid. CalcLotForATRStop() uses lossPerLot = (ATR × 1.5 / tickSize) × tickValue and divides the risk money by that. The SL is placed at entry − ATR × 1.5 (or mirror for short) and the TP at entry + ATR × 2.0 — a 1.5:2.0 ratio, i.e. an R-multiple of 1.33 against the stop. The stop distance is also floored at the symbol's SYMBOL_TRADE_STOPS_LEVEL so the broker never rejects the order for stop-distance violation. When InpCapitalCapAmount > 0 the EA treats that figure as the effective capital ceiling, plus today's realized PnL, which means a profitable day expands the sizing base on every subsequent trade until the session resets.

Trade management runs on every tick in ManageTrades(). When profit reaches one R the stop is moved to entry + 15 points (or mirror) — a break-even lock with a small offset to absorb spread. At the same R threshold, half of the position is closed if the remaining volume is at least 1.5× the broker's minimum lot, and the comment is tagged with [P] to prevent a second partial. From the same R, an ATR(14) trail follows price at a distance of one ATR; the stop is ratcheted forward only, never backward. There is no ratchet at fractional R levels — the three behaviours (BE, partial, trail) all switch on at exactly 1.0 R, which is the deliberate simplification that makes the management state easy to inspect in a backtest. A separate time-exit closes any trade that has been open for 240 bars (20 hours of M5), and RegimeOK failing on the ADX leg causes an immediate regime-change exit.

The no-trade gate combines daily and weekly realized PnL against InpDailyLossLimitPercent (5%) and InpWeeklyLossLimitPercent (10%) of effective capital, computed from the deal history stream. A capital-floor check blocks entries when the effective capital falls below InpCapitalCapFloor (50 USD by default). A cooldown activates after three consecutive losing closes — UpdateConsecutiveAfterClose() reads the last hour of deal history, increments a counter on each negative close, and blocks new entries for the next three closed bars (15 minutes on M5). A spread filter rejects entries when the symbol spread exceeds 50 points or, if SYMBOL_TRADE_STOPS_LEVEL is wider, four times that level. A secondary filter requires the ATR in points to be at least 3.5× the spread, killing low-volatility high-spread regimes that look entry-attractive but actually pay out nothing after costs. The session window is hardcoded to server hours 8 through 20, and weekend protection closes the book on Friday after 22:00 and refuses to open on Saturday or Sunday. IsNewsTime() adds a Friday-only NFP blackout between hours 12 and 16 for the first seven days of the month, blocking entries without forcing a position close.

InpDryRun defaults to true — the EA prints the full signal context, regime state, and order parameters in the log but does not send anything. The user must explicitly set it to false to take the EA live, and the on-screen tag [DRY-RUN] makes the state obvious in the Experts log. InpKillSwitch is a separate input that, when true, calls CloseAllOurPositions() on the next tick and stops opening anything. The pyramid code is present and functional but disabled by default: InpEnablePyramid = false. When enabled, it adds to the most profitable open position only, only if that position has moved at least 1.5× ATR in profit, and only up to a hard cap of 3 levels; a new pyramid entry also has to clear RegimeOK and the order margin check before it sends. The filling mode is auto-picked from the symbol's SYMBOL_FILLING_MODE mask in the order FOK > IOC > RETURN. Send retries use 40 points of deviation, with a single refresh-rates retry on REQUOTE, PRICE_OFF, or TIMEOUT; close and partial-close operations use 200 ms backoff for up to three attempts on the same retcodes plus PRICE_CHANGED; modifies use 100 ms backoff for the same set. The OnTester fitness function is (net × profitFactor) / (1 + balance-DD-%), gated by a minimum of 30 trades — anything below 30 trades returns 0 so the strategy tester can prune statistically meaningless passes.

What this EA will look like in a backtest: an XAUUSD M5 chart with a 9/21 EMA stack, Bollinger Bands at 20/2.0, an M15 trend-confirm overlay, and a relatively low trade frequency because the regime classifier and triple-confirmation rules filter most bars out. The trade count per week is small, the win rate is moderate, and most of the equity curve comes from the asymmetry of the partial-TP-at-1R exit combined with the ATR trail. The most common failure mode is the EA being glued to the QUIET or RANGE state during a transition into a trend, which is why the HTF confirm toggle is the right place to start tuning: turning it off raises trade count at the cost of lower-quality signals, and turning it on with a stricter threshold (e.g. M30 instead of M15) trades count for stability.