Pipsgrowth EX18102 TrendFollow

EX18102 is a single-indicator session-timed reversal EA. The strategy watches a 28-period Relative Strength Index on the M15 chart, looks for RSI(28) to cross the 30 (oversold) and 68 (overbought) thresholds, and treats those crossovers as buy and sell signals. The whole system is gated to the Asian session — 00:00 to 08:00 UTC — and any open position is force-closed as soon as the clock leaves that window.

The signal is a true reversal pattern, not a trend-following one. The long entry fires when the previous bar's RSI was above 30 and the current bar's RSI has dropped to or through 30. In other words, the EA buys into exhaustion: when momentum falls into deep oversold on the M15 chart, the EA enters long and waits for a mean reversion. The short entry mirrors it — RSI crosses above 68, the EA sells. Both thresholds sit slightly inside the conventional 30/70 band, which filters out a few of the marginal crossovers you would see at the textbook levels.

RSI computation lives in a single indicator handle, rsiHandle = iRSI(_Symbol, PERIOD_M15, RSIPeriod, PRICE_CLOSE). The EA reads the last three RSI values into a 3-slot buffer (rsiCurrent, rsiPrevious, rsiPrevious2) on every tick, then CheckRSICrossover() looks at the current-vs-previous bar to flag three boolean states: rsiCrossedOversold, rsiCrossedOverbought, and rsiCrossedExitLevel. Those flags feed both the entry logic and the on-chart panel.

The session filter is implemented in two places. IsAsianSession() returns true when timeStruct.hour >= 0 && timeStruct.hour < 8. When the tick falls outside that window, the EA either returns immediately (no new entries) or — if CloseOutsideSession is true — runs CloseAllTrades("Outside Asian session") to flatten any open position. There is also a sessionCloseAttempted boolean that prevents the close order from being issued more than once per non-Asian block. The same time-struct lookup drives GetCurrentSession(), which labels each hour as Asian, London (08:00-16:00 UTC), New York (13:00-21:00 UTC), or "Other" for the 21:00-00:00 gap. The two New York / London windows overlap at 13:00-16:00 UTC, but the EA's trade gating still depends only on the Asian block.

Position sizing is fixed, not risk-based. The input MaxLotSize = 0.2 is the literal volume written to trade.Buy(...) and trade.Sell(...). There is no AccountInfoDouble(ACCOUNT_BALANCE) * riskPercent / slDistance calculation anywhere in the source, and the lot size does not adapt to the SL distance when one is enabled. This is unusual for a PipsGrowth EA and is one of the most consequential design choices in the file — the dollar risk per trade is determined entirely by the broker's tick value, not by the EA.

SL and TP are both off by default (UseStopLoss = false, UseTakeProfit = false). When disabled, trade.Buy(MaxLotSize, _Symbol, currentAsk, 0, 0, ...) opens the position with zero brackets — the trade has no protective stop and no take profit at the broker. The two pip inputs StopLossPips = 5 and TakeProfitPips = 175 are still there in the parameter list, and the panel always displays them, but they are dormant unless you flip the boolean flags. A 5-pip stop on AUDUSD M15 is extremely tight — roughly the size of a single bar's range in active hours. A 175-pip target implies a 1:35 reward-to-risk structure that almost never plays out on a mean-reversion trade.

There are two real exit paths. The first is a time stop: if TimeCurrent() - positionOpenTime > MaxDuration * 3600 (340 hours by default — about 14 days), CloseAllTrades("Timeout") is called. The second is the RSI exit, gated by UseRSIExit = true and RSIExitLevel = 48. With the default 48, the EA closes a long when RSI crosses above 48 and closes a short when RSI crosses below 48. The level is set just below the 50 midline, which means longs exit on a return to neutral momentum, not on a full reversion to overbought. Shorts have to give back roughly the same distance to trigger.

Order execution goes through the standard CTrade wrapper. trade.SetDeviationInPoints(3) allows 3 points of slippage, trade.SetTypeFilling(ORDER_FILLING_IOC) requests Immediate-Or-Cancel filling, and trade.SetExpertMagicNumber(22218102) ties every order to this EA's magic. The single-position gate is the loop over PositionsTotal() that runs in OnTick — if any position exists on the symbol, no new entries are considered. There is no pyramiding, no scaling-in, and no re-entry logic.

The retry layer is a copy-paste of the standard PipsGrowth helper set: TryClose_EX18102(ulong ticket), TryClose_EX18102(string symbol), TryClosePartial_EX18102(ticket, volume), and TryModify_EX18102(ticket, sl, tp). Each loops three times with a 100-200 ms sleep between attempts, only retrying on TRADE_RETCODE_REQUOTE, TIMEOUT, PRICE_OFF, and PRICE_CHANGED. In the live execution path, only the string symbol variant of TryClose_EX18102 is actually called from CloseAllTrades() — the partial-close and modify helpers are dead code in this build.

The on-chart panel is the most visible feature of the EA. CreatePanel() builds a 200x200 black rectangle with a "RSI Reversal" title and seven data labels: RSI value, current position, spread in pips, session name, stop loss, take profit, and the most recent crossover state (None / Overbought / Oversold / Exit). UpdatePanel() rewrites the seven labels on every tick. The panel position is configurable via PanelX and PanelY (defaults 10 and 20 — top-left corner). Panel colors are also exposed as inputs (PanelBackground, PanelText).

Two non-obvious things show up if you read the source carefully. First, MaxSpread = 1000 is a default so loose it is effectively disabled — at AUDUSD's typical 0.5-2 pip spread, the filter is never the binding constraint. Second, CloseAllTrades() contains a reference to PositionGetInteger(POSITION_MAGIC) == 123456 — a leftover from the template generator. The result of that check (hasOurPositions) is never used afterwards, and all real closure uses the symbol-based path, so the wrong magic number is a no-op rather than a bug in the live behavior. But it is a sign the file came from a stock template without a final per-EA scrub.

The header block claims a "12-layer architecture" (REGIME, SIGNAL, ENTRY, CONFIRM, NO-TRADE, CAPITAL CAP, RISK, SIZING, MANAGE, EXIT, SCALING, OnTester). In the actual code only SIGNAL, ENTRY, SESSION-FILTER (a narrow slice of NO-TRADE), and EXIT are real. There is no regime classifier, no higher-timeframe confirmation, no capital cap, no risk-percentage sizing, no scaling module, and no OnTester function. Treat the 12-layer claim as marketing copy, not an implementation contract.

What to expect in backtest: a low-frequency, session-restricted mean-reversion profile on AUDUSD M15. Win rate on a 5-pip-stop / 175-pip-target asymmetry is structurally going to be poor — the SL fires constantly during ranging hours and the TP almost never hits. The realistic edge has to come from the RSI-exit path (closing at 48 instead of waiting for the 175-pip TP) and from the Asian session filter (cutting exposure before the London break). If you run this on a long enough window, the equity curve tends to drift, not to climb, and the dominant cost is broker spread on the 5-pip stops. With UseStopLoss = true and a wider stop — say 25-40 pips on M15 — the profile looks different, but you are now trading the EA's exit discipline, not the EA's signal.

For a live setup, run this on an AUDUSD M15 chart on a low-spread ECN or RAW account, with UseStopLoss = true at a sensible distance for the pair (20-30 pips, not 5), UseRSIExit = true, and UseTakeProfit = false (let the RSI exit do the work). The 100 USD minimum deposit that the input block lists is only meaningful for the fixed 0.2-lot sizing — you need enough margin to absorb a single M15 reversal without a margin call. If you want to scale up, change MaxLotSize proportionally; the EA will not adapt it for you.