Pipsgrowth EX18073 TrendFollow

EX18073 is a structural trend EA that combines a dual-EMA directional filter with a self-built support/resistance map reconstructed from recent price memory. Instead of relying on a static pivot table or a single swing high/low, the EA scans the last InpSRClusterBars closed bars (default 200), bins the close prices into 5-pip-wide histogram cells, and treats any cell that has been touched at least InpSRMinTouches times (default 3) as a price-memory level. The current ask or bid is then compared against that level — if the price is sitting within SR_PROXIMITY_PIPS (3 pips) of a populated cluster, the level is treated as live support (for longs) or live resistance (for shorts). The trade is allowed only if the level sits on the correct side: a BUY requires the price to be above the cluster, a SELL requires it to be below. The intuition is that price memory is real: levels that have been revisited multiple times in the recent past act as a magnet or a barrier, and the EA wants to fade a bounce off that level in the direction of the prevailing trend.

The trend itself is established by a fast/slow EMA consensus. The fast EMA is InpFastMA (default 25), the slow is InpSlowMA (default 50), both exponential and both applied to close. EmaSide() returns +1/-1/0 based on the closed-bar (shift 1) relationship between the two. EmaSlopeAgrees() then checks that the fast EMA has been monotonically moving in that direction over the last two closed bars — a price above the slow EMA is not enough on its own, the fast line must also be sloping the right way. The trade is rejected if the fast EMA is flat or counter-trend on the previous bar. This rules out the most common false signal: a price that has crossed above the slow EMA but is grinding sideways while the fast EMA is curling back down.

A regime classifier runs on top of the trend filter. ClassifyRegime() reads three live series — ATR(14), ADX(14), and Bollinger Bands(20, 2.0) — and produces one of five labels. STRONG_TREND requires ADX ≥ 20 and the current ATR(14) to be at or above the 30th percentile of the last 50 bars. WEAK_TREND relaxes the thresholds to ADX ≥ 14 and ATR percentile ≥ 15. COMPRESS is detected when the current Bollinger bandwidth is less than 90% of the previous bar's bandwidth. EXPAND is the opposite, with the bandwidth ratio > 1.10. Anything that doesn't fit is labelled RANGE. RegimeAllowsEntry() accepts STRONG_TREND, WEAK_TREND, EXPAND, and BREAKOUT — the EA will not open new positions during COMPRESS or RANGE regimes. The COMPRESS gate is a meaningful filter on XAUUSD M5, where the market regularly alternates between quiet consolidation (low ATR, contracting bands) and explosive breakouts, and entering during a squeeze is exactly the wrong side of the distribution.

Confirmation is layered. The first check is ConfirmHTF(), which reads a higher-timeframe EMA(50) — by default the H1 EMA via InpHTF = 8 mapped through MapTimeframeInt(). A long is confirmed only if the M5 closed bar finished above the H1 EMA(50) AND the H1 EMA has been sloping up over its last two closed bars. A short requires the mirror. The second confirmation is the S/R cluster test described above. The third is ConfirmRR(), which checks that the proposed SL/TP distances produce a reward:risk ratio of at least InpMinRR (1.5). With InpSLATR = 1.5 and InpTPATR = 3.0 the actual ratio is exactly 1:2, so the MinRR is a floor, not the operative value — but it does prevent the EA from sending a trade where the SL was widened by StopsLevelPrice() (the broker's minimum stop distance) to the point that the math no longer works.

The no-trade gate is the standard seven-layer filter. NoTradeBlocking() rejects entries if any of the following is true: InpKillSwitch is on, effective capital is below InpCapitalCapFloor ($50 by default), the current spread exceeds InpMaxSpreadPips (3.0 pips on XAUUSD means the broker is quoting less than 30 points on the typical MT5 gold feed), the day is Saturday or Sunday, the time is Friday after 20:00 server time, the current server hour is outside the InpSessionStartHour–InpSessionEndHour window (7–21 by default), or a cooldown is active. Realized PnL is checked against the InpDailyLossLimitPct (3% of effective capital) and a hard-coded weekly limit (WEEKLY_LOSS_LIMIT_PCT = 6.0).

Position sizing is risk-based. CalcLotsByRisk() takes the proposed SL distance in price, multiplies effective capital by InpRiskPercent / 100, and divides by the per-lot money-per-distance figure for the symbol. On XAUUSD with a 0.5% risk setting and a 1.5×ATR stop, a $1,000 account yields roughly 0.02 lots on a 30-pip SL. The lot is then normalized to the symbol's LotsStep and bounded by LotsMin/LotsMax. If the result is zero, the trade is skipped. Effective capital is g_Acc.Equity() minus today's realized PnL, optionally floored by InpCapitalCapAmount (a hard cap on real-money exposure) and always floored by InpCapitalCapFloor. With InpCapitalCapAmount = 0.0 (the default), the cap is effectively off and effective capital equals base equity.

Trade management runs every tick on every open position. Three rules fire in sequence. First, a partial close of PARTIAL_TP_PCT (50%) of the position when the trade has moved InpPartialTPR (1.0R) in profit. Second, a one-shot move of the stop to breakeven at InpBreakevenAtR (1.0R), with a StopsLevelPrice() safety pad to satisfy the broker's minimum stop distance. Third, an ATR trailing stop at InpTrailATR × ATR(14) (2.0× ATR by default), ratcheting forward only when the new proposed stop is at least TRAIL_STEP_ATR × ATR (0.5× ATR) better than the current one. The trail only kicks in if InpTrailATR > 0 and ATR is non-zero. The 50% partial only fires if the remaining position is at least 2× the symbol's minimum lot.

Exit logic covers three paths beyond the trailing stop and the hard TP. There is no time stop — MAX_BARS_IN_TRADE = 0 disables the bar-age exit entirely. There is a regime-change exit that closes the position only if the current regime is RANGE or COMPRESS and the position is currently in loss (profit < 0); winning positions in those regimes are left to run. There is an opposite-signal exit that closes the position the moment EmaSide() flips to the opposite direction on a closed bar. All three are wrapped in TryClose_EX18073(), a 3-attempt retry helper that sleeps 200ms on requote/timeout/price-off/price-changed.

Execution is conservative. OpenMarket() runs OrderCalcMargin first and refuses to send if the required margin exceeds free margin, returning an INSUFFICIENT_MARGIN reason. On a failed send (requote/timeout/price-changed) it refreshes the prices and retries once. The deviation is SLIPPAGE_POINTS = 30 (3 pips on a 5-digit XAUUSD feed). SetTypeFillingBySymbol() adapts the order type to whatever the broker accepts. The InpDryRun flag defaults to true — the EA will print the intended order to the log but not send it, which is the right default for a structural strategy that you want to validate in the Strategy Tester before going live. Set InpDryRun = false to enable real trading. The order comment is hardcoded to Psgrowth.com Expert_18073.

Capacity is intentionally tight. InpMaxConcurrent = 1 and InpMaxSymbolPos = 1 mean the EA holds at most one open position on the symbol at any time. There is no pyramiding and no scaling — the partial TP at 1R is the only reduction in exposure, and it leaves the second half of the position on for the trail to manage. This makes the EA a single-shot, high-conviction trader: it spends most of its time waiting, fires only when the S/R cluster test, the dual-EMA consensus, the H1 trend, and the regime classifier all line up on the same bar, and then manages the resulting position with a simple BE + partial + trail structure.

The OnTester fitness function is (netProfit × profitFactor) / (1 + balanceDrawdown), returning 0 if the total trade count is below 30. This is the standard trend-EA fitness recipe in the corpus: it rewards both raw profit and profit factor, and it penalises drawdown, but it does not directly reward the Sharpe ratio or reward smoothness. Use the standard MT5 optimization pass with this as the criterion when sweeping the EMA periods, the S/R cluster window, and the ATR regime thresholds.

What to expect in backtest: the histogram-binned S/R map is a slow-to-adapt feature, so on regime changes the cluster levels can lag the market by a few bars. The 200-bar window on M5 is roughly 16 hours of trading, which means the levels the EA is testing against were generated during the previous London and New York sessions. On XAUUSD, where the Asia-London overlap routinely produces the day's true range, this can be a feature (the level is real) or a bug (the level is stale). The combination of a single-position cap and a 1:2 hard TP means the EA will not catch very long trends — the partial + BE structure assumes you are happy banking 1R and trailing the rest. If you are looking for a multi-day trend rider, this is not the right tool. If you are looking for a clean, structural, mid-frequency trend trade on XAUUSD M5 with a clear entry, a defined stop, and a managed exit, this is.