Pipsgrowth EX18005 TrendFollow

Pipsgrowth EX18005 TrendlineCrossAlert does not invent patterns; it waits for price to cross a trendline, exactly the way a discretionary trader would draw one on a chart, then executes through the same risk and regime gates the rest of the TrendFollow family uses. The feature that separates it from a generic crossover EA is that the trendline is either computed inline from the last N closed bars using a least-squares linear regression, or read directly from chart objects the operator has drawn themselves (HLINE or OBJ_TREND). When both are active, the manually drawn line wins, so the trader can override the algorithmic baseline with their own view of where the market is breaking.

The automatic trendline is built on the fly by the function TrendlineSignal(). It pulls the last InpAutoTrendlineBars closed bars (default 50, so roughly four hours of M5 price), feeds close values and bar indices into the standard slope/intercept formula (slope = (N·Σxy − Σx·Σy) / (N·Σx² − (Σx)²), intercept = (Σy − slope·Σx) / N), and projects the line value at bar[1] and bar[2]. A bullish cross requires close2 ≤ lineAt2 and close1 > lineAt1; a bearish cross is the mirror. The function also walks the chart's OBJ_HLINE and OBJ_TREND objects, keeps the one whose value at bar[1] sits within 1% of the current close, and substitutes it as the active line — that is the manual override. If InpAutoTrendlineBars is set to 0 and InpUseManualLines is true, the regression is skipped entirely and the EA waits for the operator's own line to be broken.

Confidence is not a fixed number. It is calculated as min(100, ADX(14)[0] × 2) and floored at 10, so a quiet market with ADX around 15 yields a 30-confidence signal while a strong trend with ADX 45 yields 90. The signal only reads bar 1 and bar 2, which means it can only fire on a closed bar — no intrabar flicker triggers an entry.

Before any order is sent, DetectRegime() runs on every tick. It reads ATR(14), computes a percentile against the last 50 values, and combines the result with ADX(14) into a four-state classifier. REG_STRONG_TREND requires ADX ≥ InpADXMinTrend (20) and ATR percentile above 0.5. REG_WEAK_TREND fires when ADX ≥ 18. REG_RANGE trips when ATR percentile is below 0.25. Everything else falls into REG_CHOPPY and is hard-blocked. The indicators that feed this classifier are created in OnInit(): iATR(14), iADX(14), iBands(20, 2.0), and iMA(50) on PERIOD_H4 — note that the higher timeframe here is H4, not H1, so the trend agreement is read on a slower cycle than some of the other family members.

ConfirmEntry() then adds two more gates. The H4 EMA(50) must agree with trade direction (price above the EMA for longs, below for shorts), and the implied R:R must clear InpRRMin (1.5). At the default ATR multipliers of 1.5× for the stop and 3.0× for the target, the design ratio is 1:2, which is comfortably above the 1.5 floor.

CanTradeNow() is the no-trade gate. It refuses entry on: kill switch active, CHOPPY regime, capital floor breach (effective capital below InpCapitalCapFloor of 50 USD), spread over 50 points, server hour outside 7-20, weekends, daily realized loss at or above 3% of effective capital, weekly realized loss at or above 6%, active cooldown, or position count at the cap. The realized PnL trackers RealizedToday() and RealizedThisWeek() walk the history deals filtered to this magic only — no other EA's trades are counted in the daily or weekly ledger.

The cooldown mechanism is wired into OnTrade() rather than OnTick(). After a closing deal with negative net PnL, g_consecutiveLosses increments; once it hits InpCooldownAfterLosses (3), g_cooldownUntil is set to TimeCurrent() plus three bars of PeriodSeconds, which works out to 15 minutes on M5 or 3 hours on H1, the counter resets, and the EA logs the cooldown window. A winning trade resets the counter immediately. This is faster than the same-bar cooldown you sometimes see in cheaper EAs because it fires the moment a deal closes, not on the next bar open.

Risk and sizing are deliberately simple. CalcLots() divides EffCapital() × (InpRiskPercent / 100) (default 0.5% of equity, capped at InpCapitalCapAmount when set) by the per-lot loss at the ATR-derived stop distance, then runs the raw value through NormVol() which floors to the lot step and clamps to the symbol's min and max. Effective capital is min(account equity, InpCapitalCapAmount), so you can hard-cap a prop-firm allocation at any dollar figure without rewriting the EA.

The stop is computed at order time as ATR(14) × InpATRMultSL (1.5× by default), then the resulting distance is checked against the broker's stops level via StopsPoints() — if the distance is below the floor, the stop is widened by one point beyond the floor. The target is ATR(14) × InpATRMultTP (3.0×), giving a 1:2 R:R unless the multipliers are changed. OrderCalcMargin() pre-checks margin before the order is sent, and the trade object retries once on requote, price-changed, or timeout before giving up.

Position management runs every tick through ManageOpenPositions(). There are four exit paths, applied in this order: a 200-bar time stop, a regime-change exit (CHOPPY or RANGE closes the position outright), a one-shot break-even at 1R that moves the stop to entry plus or minus one point and only ever ratchets in the favorable direction, and a partial close of 50% at exactly 1R (so the partial fires the same tick as the break-even). The ATR trail kicks in at 1R and runs at ATR(14) × InpTrailATRMult (2.0× by default) from the current bid/ask, also only ever moving the stop favorably. Every modify and partial close is wrapped in TryModify_EX18005() and TryClosePartial_EX18005() — both retry up to 3 times on requote, timeout, price-off, or price-changed with 200ms sleeps.

There is no explicit basket target layer; the per-trade 1:2 R:R plus the partial-at-1R do the heavy lifting. There is no active news filter as code, but the NEWS_WINDOW_MIN constant is reserved (default ±30 min) for future use. The dry-run switch (InpDryRun = true by default) prints the intended order to the log without sending it, which is the safest way to validate that the trendline signal is firing where you expect before flipping to live.

The OnTester() formula is (STAT_PROFIT × STAT_PROFIT_FACTOR) / (1.0 + STAT_BALANCE_DD), returning 0 if the test has fewer than 30 trades or a degenerate PF or DD. The optimizer is therefore biased toward high-PF, low-DD, profitable configurations, and the 30-trade minimum prevents overfitting on a handful of lucky setups.

For live deployment, the trade-off to understand is that this EA does nothing on a market without a trendline break. If the regression slope is approximately zero, no cross can form and you will see long stretches of inactivity — that is by design. The EA is built for the trader who wants to be in only when a structural level gives way, not when a noisy oscillator flickers. Set InpAutoTrendlineBars higher (80 to 100) on noisier symbols like XAUUSD M5 to smooth the regression; drop it to 25-30 on H1 where the bars are fewer and the slope is more meaningful. Manual lines are useful for marking news-event highs and prior-day levels — the EA will trade the first clean break of any chart object that ends up within 1% of recent price.