Pipsgrowth EX18003 TrendFollow

Pipsgrowth EX18003 TrendFollow is a closed-bar trend follower for MetaTrader 5 whose centre of gravity is not the signal itself but the eight-state regime classifier that sits in front of it. The trade idea is the simplest possible: an EMA(10) crossing an EMA(50) on a closed bar, and the EA will only act on that one shape. Everything else in the 1,000-line source is about deciding whether the cross is real, whether the cross is worth trading, and how to take the money off the table once it is in. There are no martingale legs, no grid layers, no recovery sequences — entries are one position at a time, and a position is closed by one of five distinct exit paths. That is the whole design.

The decision pipeline in OnTick() is strictly ordered. The very first thing the EA does on a new bar is call ClassifyRegime(), which reads ADX(14) and Bollinger Band width off a 100-bar history and labels the market into one of seven states: REGIME_CHOPPY when ADX is below 18, REGIME_BREAKOUT when Bollinger width is at or above the 80th percentile, REGIME_COMPRESS when width is at or below the 20th percentile, REGIME_STRONG when ADX is at or above 25, REGIME_WEAK when ADX is at or above 14 but below 25, REGIME_RANGE otherwise, and a final REGIME_EXPAND override that fires when both the ATR and BB-width percentiles are above 70 with ADX at or above 20. The label gets cached into a global so the rest of the bar can read it. The signal itself — SignalDirection() reading two closed bars of the fast and slow EMAs and returning +1/-1/0 — runs only after the regime pass.

Before any signal is even looked at, the EA walks NoTradeBlock(). This is a single function but it is a long one: it returns true (and the bar is skipped) on any of ten guards. The capital floor is enforced first: if g_effectiveCap has dropped below InpCapitalCapFloor (50 USD by default), no new entries. Then the spread guard: if the symbol spread is more than MAX_SPREAD_MULT (3.0) times the rolling 50-tick average, the bar is skipped. Then the session window: SESSION_START_H=7 to SESSION_END_H=20 in server hour. Then the weekend-proximity block: Friday after 22:00, or any time on Saturday or Sunday. Then the daily loss limit: realised PnL for the EA's own magic on this symbol, recomputed from the deal history, cannot be worse than -3% of effective capital. Then the weekly loss limit: same logic, -6% of effective capital, week starting Monday at 00:00 server. Then the cooldown flag, the max-concurrent check, the data-quality check on bid/ask, and the broker-trade-disabled check. Any one of those returns true and the rest of the bar is dead.

If the regime is labelled and the no-trade gates are all clear, the signal can finally run. SignalDirection() reads the two most recent closed bars of the fast and slow EMAs and checks for a strict crossover — the fast must be at or below the slow on bar 2 and strictly above on bar 1, or vice versa. A wick-cross that does not close across does not count. A close that happened two bars ago does not count. The signal is then handed to ConfirmFilters(), which applies four more side-checks. The H1 EMA(50) must be on the same side of the close as the signal. The RSI(14) must be on the right side of 50 — above for longs, below for shorts. The ADX(14) on the closed bar must be at or above InpAdxMin (20 by default). And the prospective trade must have a risk-to-reward of at least InpMinRR (1.0), measured as the 3×ATR take-profit distance divided by the 2×ATR stop distance. The final confirm gate is the regime: REGIME_CHOPPY and REGIME_COMPRESS reject the trade outright. The EA logs the exact reason for any rejection (HTF-EMA-disagree, RSI-side, ADX-low, RR-low, regime-blocked) so the strategy tester output is readable.

TryOpen() does the order. The position size comes from InpRiskPercent (0.5% default) applied to g_effectiveCap, which is min(InpCapitalCapAmount, equity) when a capital cap is set and just equity otherwise. CalcLot() converts the risk budget into lots via tickValue, tickSize, and the ATR-derived SL distance, and ClampLot() snaps the result to the broker's lot step and bounds. A portfolio-heat check refuses the trade if the sum of all open R-exposures would push total risk above MAX_PORTFOLIO_HEAT (8%). A OrderCalcMargin() pre-check ensures free margin is sufficient. The order is sent with trade.Buy or trade.Sell; on TRADE_RETCODE_REQUOTE or TRADE_RETCODE_PRICE_OFF or TRADE_RETCODE_TIMEOUT, the EA refreshes the quote and retries once. Critically, InpDryRun defaults to true — the EA logs the signal but does not send the order until the operator explicitly turns live trading on. This is the default for the published build, not an oversight.

Once a position is open, ManageOpenPositions() runs every tick to do the post-entry work. The same SL_ATR_MULT=2.0 and TP_ATR_MULT=3.0 baseline used at entry is re-evaluated. Three things happen on every tick. First, at the 1R mark (BE_TRIGGER_R=1.0), the stop is moved to entry plus a 5-point buffer — break-even is a one-shot ratchet, not a perpetual one. Second, at the same 1R mark, the EA attempts a partial close of PARTIAL_FRAC=0.5 (50%), as long as the partial volume is at or above the broker minimum lot and is strictly less than the full position. Third, an ATR trail of InpAtrTrailMult (3.0×) is applied bar by bar: the new stop is current - 3×ATR for longs and current + 3×ATR for shorts, with the requirement that it advances beyond the current stop and stays below/above the live price. Three more exits are also handled: a regime-flip exit that closes any position in clear profit (R ≥ 0.5) if the regime becomes Choppy or Compress, an opposite-signal exit that fires the moment the EMA crossover direction reverses, and a momentum-decay exit (MomentumDecay()) that closes when five consecutive closed bars show strictly shrinking body size and the most recent bar is against the position. A time-based exit at InpMaxBarsInTrade (200 bars) prevents zombie trades that get stuck mid-channel.

The capital cap deserves its own paragraph. With InpCapitalCapAmount=0 (the default), the cap is disabled and the EA risks off full equity — straightforward. When the operator sets a real value, every risk calculation in the EA uses min(cap, equity) as the denominator, so CalcLot() does not scale with a runaway account. This is a prop-firm-friendly feature: set the cap to the firm-allowed balance, and the EA's risk math will silently stop growing the position size once the account is past the cap. The 50 USD InpCapitalCapFloor is the corresponding lower guard: any account drop below that value blocks all new entries. Combined with the 3% daily and 6% weekly loss limits (also expressed as a percentage of effective capital), the EA has three layered equity guards, and the weekly loss is recomputed each tick from the deal history (Monday 00:00 server time onward), not from a separate PnL feed, so the EA cannot be fooled by deposits or withdrawals that happen mid-week.

OnTester() returns the optimisation score (netProfit × profitFactor) / (1 + equityDrawdownPercent), but only if the run has at least 30 trades (TESTER_MIN_TRADES=30); otherwise it returns 0 so that low-sample backtests do not get preferred. A run with no payoff is given PF=0.001 and a no-drawdown run is given DD=0.001 as a guard against divide-by-zero. The score is a robustness-first measure: a high-profit, high-drawdown run scores lower than a moderate-profit, low-drawdown run with the same number of trades. The EA also exposes a InpKillSwitch input that, when set to true, freezes new entries while still managing open positions — useful for pausing a live instance at the broker terminal without having to remove it.

For deployment, the source header notes the EA is portable to FX majors and other metals, but the default chart is XAUUSD on M5, with M5 through H1 as the working range. The 0.5% default risk per trade and 50 USD capital floor mean the EA is comfortable on a 100 USD account — the floor is hit only at the very tail of a losing streak. The session window of 7-to-20 server hour corresponds to the London–New York overlap plus the late-Asia wind-down, which is the right liquidity window for a crossover system on gold. A retail broker with no minimum-stop-policy and a tight spread on XAUUSD is the natural home; the spread guard exists precisely to keep the EA out of toxic fills when the spread widens.