Pipsgrowth EX14015 Scalper

EX14015 EMA Fractal Scalper is a 3-EMA trend-pullback system that fires only when the slow, medium, and fast exponential averages stack in a clean sequence, then waits for a Williams Fractal pivot to confirm the reversal of a short-term dip. The EA does not chase breakouts and it does not fade ranges. It looks for an established directional bias, identifies a controlled retracement into the fast EMA, and enters on the first confirmed fractal swing that closes back on the right side of the slow average. Magic 22214015, four native indicator handles, twenty-two inputs grouped into EMA, Fractal, Trade, Stop Loss, Session, and Dashboard sections.

The trend gate is structural. The EA creates three iMA handles on the chart's current period with applied price selectable from seven modes (close by default, or open, high, low, median, typical, weighted). The check is a strict ordering: an uptrend is when EMA20 sits above EMA50 and EMA50 sits above EMA100, a downtrend is the mirror, and any other arrangement is treated as a ranging market and produces no signal. OnInit enforces this with an INIT_PARAMETERS_INCORRECT return if the user enters the periods in the wrong order. This is more than a numerical check — it is the EA's hard filter, which means a chopped, overlapping EMA stack never generates entries, no matter how many fractals print.

The entry trigger is a Williams Fractal pivot occurring on the back side of a pullback. The EA reads the iFractals handle (buffer 0 for upper fractals, buffer 1 for lower fractals) and indexes the confirmed fractal at Fractal_Offset bars back, which defaults to 2 — the standard two-bar confirmation delay. For a buy, the code requires the recent bar's low to have dipped below EMA20 (the pullback) but its close to have finished back above EMA100 (the trend is intact). It then checks that the lower-fractal buffer at the offset bar is not EMPTY and not zero. A sell mirrors this: a high above EMA20, a close below EMA100, and a non-empty upper-fractal at the offset bar. Crucially, the function tracks the time of the last signal via g_lastSignalTime and ignores any fractal that does not bring a newer timestamp, so a single valid swing cannot fire multiple times across consecutive ticks.

Stop loss placement is adaptive and pivot-aware. The CalculateSLDistance() function reads the last bar's close, high, and low plus the medium and slow EMAs. For a regular buy, the stop is the distance from the close down to EMA50 plus a 50-point buffer. The function then checks whether the bar's low came within 1.5x of EMA100 as it did to EMA50 — if so, the trade is classified as a deep pullback and the stop instead anchors to EMA100 with the same 50-point buffer. This isDeepPullback flag is propagated to the order comment (_DEEP or _REG) so the trade history records which anchor was used. The result is a stop that hugs whichever medium- or slow-EMA pivot the price actually tested, and that respects the configurable MinSL of 50 and MaxSL of 500 points. A buy's stop is therefore dynamic, but always between 0.5 and 5.0 price units on a 5-digit gold symbol.

Take profit is a fixed multiple. The EA multiplies the SL distance by RiskRewardRatio (1.5 default) and projects that same distance from the entry in the favorable direction. There is no trailing stop, no breakeven move, no partial close, and no signal-based exit — the position is held until either the broker-side SL, the broker-side TP, or a margin call ends it. The TP is attached at order placement, not managed dynamically, which means the same trade cannot benefit from a continued trend unless the trader manually intervenes. This is intentional in the design: the EA's edge is in the entry filter, not in the exit management.

Lot sizing is dual-mode. With FixedLotSize = 0.01 (non-zero) the EA returns a normalized 0.01 lot for every order. Set FixedLotSize to 0 and the EA switches to CalculateLotSize(), which converts the SL distance in points into a lot size that risks RiskPercent of the account balance. The function pulls tick value, tick size, and point from the symbol, computes a pip value, and divides risk-amount by slPips * pipValue before flooring to the lot step and clamping to the broker's min/max. The result is a risk-based sizing model that adapts to the actual stop distance — a tight 50-point SL on a $1,000 account at 1% risk produces 0.20 lots on most gold symbols, while a 500-point deep pullback produces roughly 0.02 lots. With FixedLotSize at 0.01, the same trade would still be 0.01 lots regardless.

The pre-trade gate is layered. OnTick() runs only on a new bar, then calls IsTradingAllowed() which checks MQL_TRADE_ALLOWED and TERMINAL_TRADE_ALLOWED, plus the session filter (8 to 20 server hours when enabled, with overnight support for inverted windows). The function then copies all four indicator buffers with a depth of Fractal_Offset + 5 bars, demands the bar count exceed barsNeeded + EMA_Slow_Period to ensure the slow EMA has stabilized, and checks the current spread against MaxSpreadPoints = 30 raw points. A single-position check via HasOpenPosition() runs last — if there is any open position on the symbol with the matching magic, the EA returns without scanning for new signals. There is no scaling, no pyramiding, no averaging, and no grid.

Three retry helpers are declared at the bottom of the file — TryClose_EX14015, TryClosePartial_EX14015, and TryModify_EX14015 — all wrapping the CTrade object. Each retries up to three times on REQUOTE, TIMEOUT, PRICE_OFF, and PRICE_CHANGED codes, with a 200ms sleep between attempts for closes and 100ms for modifies. None of them are actually called from OnTick() in this build, which means the order placement path goes directly through the raw MqlTradeRequest, and the retry helpers sit unused. The Request.deviation = 30 raw points is the only slippage guard, and the GetFillingMode() helper probes SYMBOL_FILLING_MODE to choose FOK, IOC, or RETURN in that order of preference.

The chart dashboard is informative but not a control surface. It draws a 220x200 rectangle label in the upper-left corner showing the live trend classification (UPTREND, DOWNTREND, or RANGING), the most recent confirmed fractal direction, the current spread against the cap, the session status, whether a position is open, and the live values of EMA20, EMA50, and EMA100. A trio of color-coded corner labels at the top of the chart identifies the three EMA lines. The dashboard updates on every tick, the EMA labels persist on bar close, and the entire overlay is removed on deinit. The header's twelve-layer claim (REGIME, SIGNAL, ENTRY, CONFIRM, NO-TRADE, CAPITAL CAP, RISK, SIZING, MANAGE, EXIT, SCALING, OnTester) is a marketing summary; the wired layers are trend regime, fractal signal, entry execution, spread/session/position confirmation, risk-percent lot sizing, and the SL/TP bracket. There is no capital cap layer that closes trades on equity drop, no OnTester custom metric, no signal-based exit, and no scaling logic.

The right account is a small one. The minimum deposit of $100 is workable only because FixedLotSize defaults to 0.01 — a sub-fractional position that survives a 50-point stop with about $0.50 of risk on a typical gold symbol. Running on risk-percent mode at 1% on $100 with a 50-point stop still produces near-min lots, but the same setting on $1,000 with a 500-point deep pullback stop produces a more meaningful position. A clean 5-digit gold broker with 30-point average spread during the 8-to-20 server window is the working baseline. The EA is server-time aware, not session-name aware, so the user must align SessionStartHour and SessionEndHour with the broker's GMT offset for the intended liquidity window — for many gold brokers this is 8-to-20 GMT to capture London and New York without the dead Asian middle.

What to expect in backtest is a low-frequency trend-pullback model. The EA averages far fewer than one trade per day on a typical gold M1 chart because fractals only print at confirmed swing points and the EMA stack filter eliminates most of those. Expect a string of small losses when the stack chops, punctuated by 1.5R winners when a clean pullback prints inside a trending day. There is no recovery mode, no martingale, and no basket TP — a bad run is a bad run, and the only safety net is the spread filter, the session filter, and the SL. The dashboard will show RANGING for most of any range day, which is the design behaving as intended: it does not trade what it cannot classify.