Pipsgrowth EX09022 MeanReversion

EX09022 is built around a deliberately small idea: pull a fast simple moving average and a slow simple moving average off the same close-price series, draw an ATR-sized band around the slow average, and let either of two independent conditions open a trade. The first condition is a classic fast-over-slow crossover. The second is a mean-reversion trigger that fires when price has wandered more than a fixed ATR multiple away from the slow MA. Both feed into the same buy and sell booleans through an OR, so the EA treats crossover and reversion as equivalent triggers rather than competing systems. This is unusual inside the EX09 family: most of its siblings pick one direction (pure trend, pure reversion, or an explicit regime switch). EX09022 simply merges the two into a single entry model and lets the next bar decide which one fired.

The indicator stack is three handles and nothing else. A 20-period SMA on close, a 50-period SMA on close, and a 14-period ATR. There is no RSI, no Bollinger band, no ADX, no KAMA, no Donchian. The two MAs are the same iMA(_Symbol, PERIOD_CURRENT, N, 0, MODE_SMA, PRICE_CLOSE) call with the period swapped, which means the EA relies on the same averaging method and the same applied price for both lines and never mixes an EMA into a SMA pair. The ATR uses default iATR(_Symbol, PERIOD_CURRENT, 14) and is the only volatility input the strategy actually consumes. Everything else — the multiplier, the SL/TP distances, the lot size — is a fixed input. The header advertises 12 layers (REGIME, SIGNAL, ENTRY, CONFIRM, NO-TRADE, CAPITAL CAP, RISK, SIZING, MANAGE, EXIT, SCALING, OnTester), but in practice only SIGNAL, ENTRY, NO-TRADE, CAPITAL CAP, RISK, SIZING, EXIT and OnTester carry code. There is no regime classifier, no confirmation stage, no in-trade management (no trailing stop, no break-even, no partial close), and no scaling layer. Calling it a 12-layer stack is aspirational; it is closer to an 8-layer stack with four empty seats.

Signal construction lives in OnTick(). The function pulls three bars of the fast MA, three bars of the slow MA, and two bars of the ATR with CopyBuffer. A trend buy fires when prev_fast_ma <= prev_slow_ma && current_fast_ma > current_slow_ma; a trend sell is the exact mirror. A reversion buy fires when current_price < current_slow_ma - current_atr * ATR_Multiplier; a reversion sell fires when current_price > current_slow_ma + current_atr * ATR_Multiplier. The default ATR_Multiplier = 2.0 means price must extend two full ATR(14) units away from the slow MA before the reversion arm wakes up, which on a 5-minute XAUUSD chart with a 14-bar ATR of around 200-300 points translates to roughly $20-$30 of one-sided extension. The two booleans are merged as buy_signal = trend_buy_signal || reversion_buy_signal and the same for sell, so a single bar can in theory satisfy both arms at once. The position gate is simple: has_open_positions iterates PositionsTotal() looking for any position on _Symbol with the EA's magic, and the trade block is skipped entirely while one is open. No averaging-in, no grid, no pyramid - EX09022 is a one-position-per-symbol EA on the live side.

Order construction goes through a CTradeValidator class rather than a raw OrderSend. The class reads SYMBOL_VOLUME_MIN/MAX/STEP, SYMBOL_POINT, SYMBOL_DIGITS, and SYMBOL_TRADE_STOPS_LEVEL on init, normalizes the requested LotSize = 0.1 input to the broker's step, and checks free margin against OrderCalcMargin with a 15% buffer. If margin is insufficient the class steps the lot down by 25% repeatedly until it either finds a workable volume or hits the broker's minimum and returns 0. SL and TP are first computed as entry_price ± SL_Points/TP_Points * GetPoint() and then re-validated by ValidateStopLoss and ValidateTakeProfit against the symbol's stops-level, with a 20% additional safety buffer on top of whatever the broker reports. The defaults are SL_Points = 500 and TP_Points = 1000, which on XAUUSD with a 0.01 point is $5 of stop and $10 of target, a clean 1:2 R:R. If the broker refuses the requested level the validator slides the SL or TP to the closest valid price rather than failing the trade, which is why some live fills show stops tighter than 500 points when a broker sets a high stops-level.

Retry is double-layered. The outer ExecuteTrade function loops three times, refreshing the price from the validator between attempts and sleeping 100 ms between attempts. The inner TryOrderSend_EX09022 also loops three times and sleeps 200 ms, but only on the four transient retcodes (REQUOTE, TIMEOUT, PRICE_OFF, PRICE_CHANGED); on any other failure it breaks immediately. The two layers combined mean a real trade can absorb up to nine send attempts before the EA gives up, with delays of 100/200/200/200/100/200/200/200 ms between them. The deviation is hardcoded at 10 points; there is no input for it. A separate safety-trade routine fires once at the start of any tester run with magic 999999 to confirm the broker path works before the EA starts logging its own trades.

Risk management is wrapped in IsSafeToTrade_EX09022 and is an 8-condition AND. The conditions, in order, are: capital cap (InpCapAmount reached, stop), capital floor (InpCapFloor = 50 below the equity, stop), initial-equity floor (95% of the balance at init, stop), daily loss limit (3% of effective capital, stop), consecutive-loss cooldown (3 losses in a row triggers a 30-minute pause via g_cooldownUntil), max trades per day (50), market-open (no Saturday, no Sunday before 22 GMT, no Friday after 21 GMT), and an active session (London 7-16 GMT or New York 12-21 GMT, with an optional Asia-avoid 0-7 GMT block). An optional InpNewsFilter adds a ±15-minute time-based blackout around the London and NY open hours - the time is checked against the schedule, not against an MQL5 news feed, so it is a proxy filter, not a real economic-calendar filter. The OnTradeTransaction handler keeps the daily counters honest: it listens for TRADE_TRANSACTION_DEAL_ADD events, ignores anything whose magic does not match, sums profit+swap+commission into g_realizedToday, and increments g_consecLosses on a loss or resets it on a win.

The OnTester formula is the EX09 family default: n * p / (1 + d), where n is net profit, p is profit factor, and d is relative equity drawdown percent. The function returns 0 if there are fewer than 30 trades or if either p or d is non-positive, so single-digit-trade optimizations get discarded automatically. The cap layer (InpCapAmount/InpCapFloor) is genuinely optional - setting InpCapAmount = 0 disables the cap entirely and the EA treats the full account equity as its capital, with InpCapFloor = 50 as the only remaining floor.

What to expect on a backtest: EX09022 on XAUUSD M5 with the default 2.0 ATR multiplier is a low-trade-frequency EA. Crossover signals on a 20/50 SMA pair fire a handful of times per week on metals, and reversion triggers require a 2-ATR extension which is a meaningful move on M5. A six-month backtest typically lands somewhere between 30 and 80 trades depending on volatility regime. The 1:2 R:R bias means winners are larger than losers on average, so a 45-50% win rate can still produce a positive equity curve as long as the EA is not stopped out by the 95% equity floor or the 3-loss cooldown during a chop regime. The lack of trailing and break-even is a deliberate trade-off: the EA never locks profit, never moves to break-even, and never scales into a winner, so a trend move that runs past the 2-ATR target simply exits at the TP and the EA re-arms. Use a low-spread ECN broker for the metals pair so the SL_Points distance is not eaten by spread on every fill, and respect the 95% equity floor by running on a balance that gives the strategy at least a few weeks of room before the daily-loss limit starts to constrain it.