Pipsgrowth EX18075 TrendFollow

Pipsgrowth EX18075 TrendFollow waits for an EMA(9) / EMA(21) crossover on the most recent closed bar, then asks an inline SuperTrend — ATR(10) multiplied by 3.0 against the HL2 midpoint — to confirm the direction. If the fast line crossed up and SuperTrend reads +1, the EA looks for a long; cross down with SuperTrend at -1 sets up a short. That two-condition agreement is the heart of the engine: a single EMA cross without SuperTrend backing produces no signal at all.

Layer two is the regime classifier inside UpdateRegime(). It pulls ATR over a 100-bar lookback to compute a percentile rank, fetches ADX(14), and reads Bollinger band width from a 20-period, 2.0-deviation envelope. The eight possible states (CHOPPY, COMPRESS, RANGE, WEAK_TREND, STRONG_TREND, EXPAND, BREAKOUT) are then pruned by the gating rule that follows: CHOPPY (ADX<15), COMPRESS (band width below 0.0005), and RANGE all return false from UpdateRegime(), so no signal ever leaves that function when the market is mean-reverting or low-energy. The EA only trades STRONG_TREND, WEAK_TREND, EXPAND, and BREAKOUT — its way of saying it is not interested in flat or contracting conditions.

If a signal survives the regime filter, it still has to clear ConfirmPasses(), a 2-of-4 voting block. The four votes are: the close of the last closed bar must agree with the higher-timeframe EMA(50) direction (configurable through InpHTF and InpHTFEMA, default H1); the close must sit inside the Bollinger envelope (so the entry is not chasing a candle that has already blown through a band); the candle's own close-to-close direction must agree with the signal; and the current spread must be no more than 1.5× the rolling 10-bar average spread. Two of those four must be true, otherwise the trade is skipped. The source header mentions an RSI filter, but the implementation replaces it with the Bollinger-position check (the inline comment in ComputeSignal notes "use BB position as proxy"), so a position is rejected whenever the close is outside the bands.

The no-trade gate in NoTradePasses() is the safety net. It checks the kill switch, the cooldown timer that arms after InpCooldownLosses consecutive losses (default 3, locks for one hour), the daily 3% and weekly 6% loss caps in dollar terms, the 35-point spread cap, the 07:00–21:00 server-time session, the 30-minute news block around midnight (plus Friday 22:00 onwards and the full weekend), and the broker margin mode (only hedging or netting accounts are accepted). On top of that, EntryPasses() adds a per-bar duplicate guard so the EA cannot fire twice on the same closed bar, and a max-concurrent check of 3 positions.

Sizing uses fixed-fractional risk off the effective capital. With InpRiskPercent at 0.5, the EA risks half a percent of equity on each new position, converts the ATR(10)-based stop distance into ticks, divides the risk dollar amount by the per-tick dollar value, and rounds the result down to the broker's lot step. The capital cap (InpCapitalCapAmount, default 0 = off) provides an additional ceiling when the trader wants to risk only a slice of the account; the InpCapitalCapFloor of 50 dollars then blocks any new entries once equity drops below that level. Sizing respects the broker's minimum lot, and the OrderCalcMargin pre-check refuses the send if free margin is insufficient.

Exits are layered. ManageExits() runs on every tick. For each open position it computes the current R-multiple (price progress divided by the initial stop distance), then applies the partial-TP rule (50% off at 1R by default), the break-even move (SL to open price at 1R), and the ATR trail (2.0× ATR behind price, forward-only ratchet, also gated on R >= 1). A 200-bar time stop forces a close if the trade has been open through that many bars of the working timeframe (about 16.7 hours on M5). On top of those rules, the opposite-signal exit closes any position whose direction disagrees with the freshest signal, and the regime-change exit wipes the book when the market slides into CHOPPY or RANGE.

Position management has a built-in retry layer. TryClose_EX18075, TryClosePartial_EX18075, and TryModify_EX18075 each loop up to three times with a 200 ms sleep on requote, timeout, price-off, or price-changed responses. PositionOpen itself gets a single additional retry on the same transient codes. The order comment is hardcoded to Psgrowth.com Expert_18075, the magic is 22218075, and InpDryRun defaults to true so the EA can be paper-tested before any live capital is exposed. OnTester() returns (net * profit_factor) / (1 + relative_drawdown_percent) with a 30-trade floor — a single composite number that weights return against drawdown.

Pyramiding is wired but switched off by default. With InpPyramidEnabled set to true, the EA allows up to 3 additional entries (hard cap 5) spaced by InpPyramidStepATR × ATR(10); with it off, OpenEntry() refuses any entry while another position from this magic is still open. The result is a single-position-at-a-time default that is consistent with the EA's risk cap, and a configurable scaling mode for users who want to lean into trends.

This is a strategy for traders who trust trend filters and want every other decision automated. The risk and confirmation parameters are tight enough that the EA is unlikely to be the source of drawdown on its own — the failure modes (whipsaw in the WEAK_TREND regime, regime mis-classification, slippage on a thin pair, broker requotes during the news window) are upstream of the code and need to be tested in a backtester with realistic spread assumptions and a 30-trade minimum before the strategy is trusted with live capital.