Pipsgrowth EX18008 TrendFollow

Pipsgrowth EX18008 is a USDJPY-tuned trend rider that pairs an hourly triple-EMA stack with a fifteen-minute Hull-MA trigger, then layers in a seven-state market regime, capital-allocation cap, and a distinctive protected-pyramid gate before it ever sends a single order. The EA's brief in the source header is explicit: H1 directional bias from a 21/50/200 EMA stack, M15 inline Hull-MA slope cross on the closed bar, and ADX must clear 22 for the signal to even exist. The full architecture is split across twelve named layers — REGIME, SIGNAL, ENTRY, CONFIRM, NO-TRADE, CAPITAL CAP, RISK, SIZING, MANAGE, EXIT, SCALING, OnTester — and every layer is wired through explicit, named functions rather than being collapsed into a single OnTick body.

The H1 stack does the heavy lifting on direction. Three EMAs run on the hourly chart (periods 21, 50, 200 by default) and the EMAStackDir() function returns a tri-state label — STACK_BULL when fast > medium > slow on the closed bar and the current bar is not retracing through either, STACK_BEAR when the inverse holds, otherwise STACK_FLAT. There is no fractional or weighted read; the stack either is bullish or it isn't, which removes a large class of mid-cross whipsaw trades. To trigger a real signal, the M15 Hull-MA must confirm: HullDir() computes an inline Hull-MA from a raw close copy (Hull = 2*WMA(n/2) − WMA(n)) over the user-defined period (default 34) and asks whether the last three values are monotonically rising, monotonically falling, or flat. A long signal only fires when the H1 stack is bullish AND the M15 Hull slope is rising on the closed bar; a short signal only fires when both flip bearish. The ATR(14) on the working timeframe then feeds a confidence score — 50 baseline plus up to 50 from price-to-Hull distance scaled by ATR and from ADX above its 22 threshold — capped at 100.

That signal still has to pass a confirm layer before it becomes an order. ConfirmBuy() and ConfirmSell() demand that bid agree with the H1 50-EMA: long needs bid above the 50-EMA and either the 50-EMA still rising or bid above its current value; short needs the inverse. The confirm step also requires RSI(14) to sit in a trend band (45–78 for longs, 22–55 for shorts), so RSI outright overbought blocks new longs and outright oversold blocks new shorts. A minimum reward-to-risk filter of 1.2 (set by InpMinRR) is enforced right after the SL/TP distances are computed from ATR. With the default SL multiplier of 2.0 and TP multiplier of 3.0, the planned R:R is 1.5, comfortably above the 1.2 floor.

Regime classification is the safety net under the signal stack. UpdateRegime() runs every tick from a 100-bar ATR history plus the live ADX(14) and Bollinger(20, 2.0) width, and emits one of seven states: STRONG_TREND (ADX ≥ 30), WEAK_TREND (ADX ≥ 22), BREAKOUT (ATR percentile > 80%), EXPAND (BB width percentile > 80% AND ATR percentile > 80%), RANGE (BB width percentile < 35%), COMPRESS (BB width percentile < 20%), and CHOPPY (everything else). RegimeAllowsEntry() returns false for RANGE, COMPRESS, and CHOPPY — three out of seven — which is roughly 43% of bars in a typical FX session. When the regime hard-blooms into CHOPPY, RANGE, or COMPRESS while a position is open and in profit, ManageOpen() closes the trade on the next tick.

Risk is fixed-fractional. CalcLots() reads the effective capital cap (the minimum of InpCapAmount and current equity, so 0.0 disables the cap and lets equity drive sizing) and computes lots = cap × 0.5% / (slTicks × tickValue), then ClampVolume() snaps to the broker's volume step and clamps to the symbol's volMin/volMax. A capital floor of 50 USD blocks all entries when the cap dips below it. Daily and weekly loss limits are evaluated against the cap (3% / 6% defaults), realised PnL is summed by SumRealized() from the deal history, and a 3-loss-in-a-row trigger sets a six-hour cooldown via g_cooldownUntil. NoTradeGates() also enforces the 0–21 server-hour session, max spread of 30% of ATR(14), the regime allow-list, the max-positions cap of 5, and the broker trade-mode check. The kill-switch input flattens and halts if flipped on.

The protected-pyramid gate is the standout design choice. With InpPyramidEnable off by default, EX18008 trades a single position per side; with it on, the EA can scale up to five adds per side, but the AllBuysProtected() / AllSellsProtected() functions refuse the new add unless every existing position on that side already has its stop moved to entry-plus-buffer in profit (the buffer is 0.05 × ATR by default). The new add must also be at least 0.5 × ATR beyond the previous entry price and the lot has to clear the 0.5% risk cap. This is meaningfully stricter than a plain pyramid — it refuses to add a fresh position into a portfolio of unprotected, underwater entries.

Position management runs on every tick. ManageOpen() reads the live Hull direction and closes the position (in profit only) if the M15 Hull has flipped to the opposite slope. A stochastic %K cross from overbought 80 closes longs; a cross from oversold 20 closes shorts. A volatility-spike detector sums the last 20 ATR values and force-closes any in-profit position the moment live ATR exceeds 2.6 × that average. For trades that survive those hard exits, ManageOpen() ratchets the stop: at +1.0 × ATR it pulls to entry + 0.1 × ATR (a one-shot break-even with a small buffer), and at +1.5 × ATR it starts an ATR trail at a distance of 1.0 × ATR. At 50% of the planned TP distance, the EA partial-closes half the position via PositionClosePartial; the remainder runs to TP or the trailing stop. The opposite-signal flag inverts on the close-bar Hull flip, and on a Friday 20:00 server-time check the EA flattens every position to avoid weekend gap risk. Slippage is capped at 30 points and every send goes through a three-attempt retry loop with 200 ms (close) or 100 ms (modify) sleeps to handle requotes and price-off returns.

For the optimizer, OnTester() returns (netProfit × max(0.01, profitFactor)) / (1 + balanceDrawdownPercent) when there are at least 30 trades; this rewards raw profit and profit factor while still penalising the depth of the drawdown. The default file ships with InpDryRun = true, which is a useful self-test mode — every buy and sell logs its would-be ticket, price, SL, and TP without actually sending, so you can confirm signal flow on a demo chart before flipping the live switch.

EX18008 was tuned on USDJPY but the implementation is broker-portable. The header explicitly notes it works on FX majors and metals and is portable across M5, M15, and H1. For a forward test, expect a relatively low trade count per week (the regime block plus the ADX threshold plus the H1+Hull agreement makes this a selective strategy, not a frequent-fire one). On a typical Tokyo and London overlap you should see the most setups, with USDJPY offering the cleanest H1-stack swings. Risk the EA at 0.5% per trade on a small account to confirm drawdown tolerance, then consider raising the risk percent only after a meaningful forward sample size of 100+ trades.