Pipsgrowth EX18025 TrendFollow

Pipsgrowth EX18025 TrendRider is a single-pair gold EA that trades a self-built Kaufman Adaptive Moving Average. There is no Pine-port, no foreign indicator buffer, no copy-pasted MT4 port: the RefreshKAMA() function in the .mq5 reconstructs the KAMA series from raw close prices on every new bar, with an Efficiency Ratio lookback of InpKAMAPeriod = 14 bars, a fast smoothing constant of 2 (≈ EMA-1) and a slow smoothing constant of 30 (≈ EMA-29). The Efficiency Ratio — change over 14 bars divided by the sum of 14 one-bar absolute changes — sets the smoothing coefficient at every step, so the curve whipsaws tightly in directional runs and flattens into a near-horizontal line in chop. That adaptive behaviour is the entire reason the EA exists: a fixed EMA lags the wrong way in ranging gold, and a fixed fast EMA chops the account in ranging gold, and KAMA is the textbook compromise.

From that KAMA series the engine derives a per-bar signal in RefreshKAMA() itself, not in a separate function. For each of the last SIGNAL_PERSISTENCE + 1 = 4 closed bars it computes three numbers: kSlope (KAMA minus KAMA three bars back, set by #define SLOPE_BARS 3), pricePos (close minus KAMA) and priceChg (bar-over-bar close change). These feed a four-quadrant lookup that returns a signed score from -1.0 to +1.0 — full credit when slope, price-side and bar-change all agree, partial credit when only two of the three align, and a small counter-signal otherwise. The scores are clamped to [-1, +1] and stored in g_kamaSignal[].

AggregateSignal() then takes a decay-weighted mean across the last 3 bars (SIGNAL_PERSISTENCE = 3, SIGNAL_DECAY = 0.8), giving the most recent bar a weight of 1.0, the bar before 0.8, and the bar before that 0.64. The result is scaled to -100..+100, and TryEntry() only fires when that aggregate clears InpSignalThreshold = 70 in the bull direction or -70 in the bear direction. The 70% bar is a deliberate filter: a single KAMA tick is not enough, two are not enough, three aligned bars in a row with the most recent weighted most heavily are what the EA treats as a real move.

No signal can fire unless the regime classifier says the market is tradeable. UpdateRegime() reads ADX(14), ATR(14) and Bollinger Bands(20, 2.0) on the current chart and maps them to one of seven states. ADX above 30 puts the chart in REGIME_STRONG_TREND; ADX between 20 and 30 yields REGIME_WEAK_TREND; ADX below 20 with high ATR-percentile and rising BB width gives REGIME_BREAKOUT; high ATR-percentile alone gives REGIME_EXPAND; tight BB and low ATR-percentile gives REGIME_COMPRESS; low ATR-percentile with normal width yields REGIME_RANGE; and the catch-all is REGIME_CHOPPY. The NoTrade() gate then hard-blocks entries in REGIME_CHOPPY and REGIME_RANGE — a trend-rider that has no business in a sideways tape. REGIME_CHOPPY is also a per-tick exit trigger inside ManageExits(): as soon as the regime classifier flips the chart to choppy, every open position is force-closed. That is the EA's own way of saying "the conditions I was designed for no longer hold."

Confirmation is a separate two-factor gate. ConfirmEntry() reads the higher-timeframe EMA(50) and EMA(200) — InpHTF is set to 8, which is outside the documented 1–7 range, so MapTimeframeInt() falls through to its default: branch and resolves to PERIOD_H1. (Anyone reading the source expecting W1 or MN will get the standard H1 instead.) The function refuses a long if the H1 fast EMA is below the slow EMA, and refuses a short in the mirror case. The second factor is the closed bar's close against KAMA: a buy needs the close to be above KAMA(1), a sell needs it below. Both factors must clear before TryEntry() will size a position.

The size of every position is set by CalcLots(), which divides the dollar risk budget (0.5% of effective capital by default, via InpRiskPercent) by the per-lot loss implied by the broker's tick size and tick value at the ATR-based stop distance. Two pre-send safety checks sit underneath: OrderCalcMargin confirms the broker will accept the order, and the result is compared against AccountInfo::FreeMargin(). A CapitalCapAmount input lets a deployer cap the equity the EA treats as its own (set to 0.0 by default, which means it trades the full account balance), with a $50 floor below which new entries are blocked.

Stop and target are ATR-based and ATR-only. InpATRSLMult = 2.0 puts the stop two ATR(14) points away from the entry; InpATRTPMult = 3.0 puts the target three ATR(14) points away, which is exactly 1:1.5 reward-to-risk. The pre-trade gate refuses to send an order if that ratio would fall below MIN_RR = 1.5, so a sudden volatility crush cannot silently degrade the structure. Once in a trade, ManageExits() runs four parallel ratchets: (1) an ATR trail at InpATRTrailMult = 2.5, which only ratchets forward and only past the entry price — the stop never goes backwards and never goes into the loss zone; (2) a break-even at 1R that pulls the stop to entry plus a tiny offset; (3) a 50% partial close at 1.5R; and (4) a hard time exit after MAX_BARS_IN_TRADE = 120 bars. Opposite-signal flips also force-close: a long is closed the moment the aggregated KAMA signal crosses -70, and vice versa.

Risk scaffolding sits around all of this. InpDailyLossLimitPercent = 3.0 blocks the EA from opening new positions once realized PnL for the day is below -3% of effective capital. A weekly cap of WEEKLY_LOSS_LIMIT_PCT = 9.0% does the same on a rolling seven-day window. A counter inside UpdateLossStreak() walks the deal history; once it sees a losing close, it forces a COOLDOWN_BARS_AFTER_LOSS = 5 bar pause before the next entry. The session filter is hard-coded via #define SESSION_START_HOUR 7 and #define SESSION_END_HOUR 21 — London open through NY close in server time — and NoTrade() also refuses to trade on Saturday and Sunday. A spread gate rejects entries when the current spread exceeds InpMaxSpreadPoints = 50 points. The InpKillSwitch input flips the EA into a close-only mode that force-closes every open position and refuses to open new ones until toggled back. InpDryRun = true by default prints every would-be fill to the journal but does not actually send orders, which is the safe way to validate the EA on a fresh broker.

Filling mode is auto-detected from the symbol: FOK if the broker advertises it, otherwise IOC, otherwise the legacy RETURN. Every send goes through a 3-attempt retry loop in TryClose_EX18025(), TryClosePartial_EX18025() and TryModify_EX18025(), with a 200 ms sleep on requote, timeout, price-changed and price-off responses. The MT5 Strategy Tester's optimization criterion is custom: OnTester() returns (netProfit * profitFactor) / (1.0 + maxBalanceDrawdown), gated to a 30-trade minimum and a positive drawdown denominator — meaning the optimizer prefers a smooth equity curve over a high-return one with deep drawdowns.

EX18025 is, in practical terms, a meta-trend-follower. The KAMA is the trend filter, the persistence-and-decay aggregator is the patience filter, the regime classifier is the volatility filter, the H1 EMA pair is the structural filter, and the ATR-based stops plus BE-plus-partial-plus-trail is the trade-management filter. Five independent reasons all have to agree before a ticket goes to the broker, and any one of them reverting kills the trade. That is the design choice that separates this EA from a simple moving-average cross: a single signal can be wrong, but five filters agreeing and then disagreeing is rare, and the EA is built around that rarity.