Pipsgrowth EX18100 TrendFollow

EX18100 is a single-indicator reversion EA that builds every decision around one oscillator — a 14-period RSI — and a deliberately asymmetric threshold architecture. There is no multi-confirmation stack, no regime classifier, no ATR filter, and no trend overlay: the EA reads RSI on a closed bar, compares it against extreme oversold and overbought zones, and either opens a position or walks away until the next bar. The name on the slug says TrendFollow and the source header talks about a 12-layer stack, but the actual code in OnInit, OnTick, UpdateRSI, CheckEntrySignals, and CheckExistingPosition is a clean four-step pipeline: read the oscillator, decide, execute, manage. Everything else mentioned in the header block is aspiration, not implementation.

The only indicator handle created in OnInit is rsi_handle = iRSI(_Symbol, g_TimeFrame, RSI_Period, g_RSI_Applied_Price) with RSI_Period=14, RSI_Applied_Price=1 mapping to PRICE_CLOSE, and g_TimeFrame derived from the integer input TimeFrame via the MapTimeframeInt switch (1=M1, 2=M5, 3=M15, 4=M30, 5=H1, 6=H4, 7=D1, default 7→D1). UpdateRSI then copies three values into the rsi_buffer array with ArraySetAsSeries at true, populating rsi_two_bars_ago (index 2), rsi_prev (index 1), and rsi_current (index 0). The crossover test in CheckEntrySignals uses bar-2 against bar-1, not bar-1 against bar-0 — a deliberate choice that means the entry decision is locked in once bar-1 closes, and the still-forming current bar cannot flip the signal. That is the EA's only anti-repaint guard.

The entry thresholds are pushed to extremes: RSI_Oversold=10 and RSI_Overbought=77. A long fires when rsi_two_bars_ago ≤ 10 AND rsi_prev > 10, which means RSI had to be sitting in deep oversold two bars ago and has now lifted off the floor by the close of the previous bar. A short fires on the mirror condition: rsi_two_bars_ago ≥ 77 AND rsi_prev < 77. These are not mean-reversion signals to the midline — they are exhaustion signals at the tail. With RSI_Overbought set to 77 rather than the textbook 70, the EA is filtering for the kind of blow-off move that has historically marked local tops in trending gold, and RSI_Oversold=10 catches only the deepest capitulation wicks. The trade frequency at default settings is therefore low — possibly a handful of signals per week on XAUUSD M15, fewer on D1.

The exit logic in CheckExistingPosition has two paths. Path one is the target hit. For a long, the position closes when rsi_current ≥ RSI_Target_Buy=27; for a short, when rsi_current ≤ RSI_Target_Sell=43. Note the asymmetry: the long target sits 17 RSI points above the entry trigger (10→27), while the short target sits 34 points below (77→43). In practice that means shorts close roughly twice as far away from the trigger as longs do in RSI terms, but RSI is bounded, so the buy side frequently closes well before price has retraced to the entry candle. Path two is the bars-against timeout. While a long is open, if rsi_current < RSI_Oversold (10), the EA starts an integer counter at 1, increments it by 1 each bar the condition holds, and closes the position once bars_against_count ≥ BarsToWait=14. The moment RSI climbs back above 10 the counter resets to zero — the EA does not punish temporary dips, only sustained ones. The short path mirrors this against RSI_Overbought (77). BarsToWait=14 is a long lookback; it gives the trade room to breathe through a 14-bar shake-out before the EA admits the thesis was wrong.

The position management is intentionally minimal. The EA does not place a stop-loss or take-profit at open — both trade.Buy and trade.Sell pass 0,0 as SL and TP. There is no trailing stop, no break-even function, no partial close, no pyramiding, and no hedge. The single-position flag (position_open bool) is the only sizing gate. LotSize=0.1 is hardcoded as a fixed lot input with no risk% calculation, no margin check beyond what CTrade does internally, and no scaling across symbols. The CTrade object is configured with SetExpertMagicNumber(22218100), SetDeviationInPoints(Slippage=3), and SetTypeFilling(ORDER_FILLING_FOK) — FOK is unusual; most modern EAs use the auto-fill policy or IOC. FOK means the broker must fill the entire order immediately or reject; for gold and majors on a fast ECN this is fine, but on a market-execution broker that requotes, every entry will fail outright. There are three retry helpers — TryClose_EX18100, TryClosePartial_EX18100, TryModify_EX18100 — but only TryClose_EX18100 is actually wired into the live code path; the other two are dead code that no function calls.

The OnTick loop is bar-paced, not tick-paced. It first checks that Bars(_Symbol, g_TimeFrame) ≥ RSI_Period + 2 (a sanity guard so the iRSI buffer has enough data), then reads the current bar time and bails out if it matches last_bar_time. Only on a new bar does it call UpdateRSI, CheckExistingPosition, and then — if no position is open — CheckEntrySignals. The TRACE prints in the source are debug leftovers and do not affect execution. The source also contains four duplicated copies of MapTimeframeInt and four of MapAppliedPriceInt — an auto-template artifact from the EA generator, harmless because the MQL5 compiler absorbs the duplicates and the runtime behaviour is identical to having just one copy. The first occurrence wins for initialization; the later ones are inert.

What the EA does not do is worth being explicit about. There is no InpMaxDDPct drawdown kill switch, no MinEquityPercent equity floor, no MaxConsecLosses cooldown, no MaxDailyLossPercent, no news filter, no session filter (it trades whenever the market is open, including the Asian overnight where gold liquidity thins out), no spread cap (so a 50-pip gold spread during a rollover can still trigger an entry), no HTF bias, no scaling, no breakeven, no partial close, no OnTester fitness function, no chart panel. Every defensive concept the parent family typically inherits has been stripped out, leaving a bare-bones oscillator reversal system.

In backtesting, the expected profile is a low-trade-count equity curve with a modest win rate, occasional larger wins when a deep oversold/overbought reading reverses cleanly to the target, and a long tail of bars-against timeouts on trades that never quite get there. The asymmetric target zones mean that when the EA does catch a reversal, the short side tends to give a better reward-to-risk in RSI terms, but in price terms that depends entirely on where gold is ranging. The two-bar-ago vs previous-bar crossover logic means a trader watching the chart in real time can see the signal one bar before the EA acts, which is useful for manual intervention but also means news-driven spikes that fail to form a closed bar above the threshold do not trigger entries. Setting RSI_Overbought lower (e.g., 70) and RSI_Oversold higher (e.g., 30) would make the EA more active at the cost of noisier signals; tightening RSI_Target_Buy down to 20 would lock in smaller but more frequent wins. The current defaults are tuned for a patient operator who wants a few high-conviction reversion trades a week, not a high-frequency scalper.