Pipsgrowth EX09019 MeanReversion

Pipsgrowth EX09019 BreakRevertPro fits three statistical probability distributions to recent M1 close prices — Weibull, Poisson, and Exponential — and reads their probability values against two thresholds to decide whether the current print is a breakout candidate or a mean-reversion candidate. Nothing in the EX09 family looks like it. EX09001–EX09006 trade on composite RSI/EMA/CCI/Stoch/candlestick math, EX09013 and EX09015 ride Bollinger band touches, and EX09016 is a stripped-down stochastic pullback. EX09019 is the only one that asks a quantitative question — does the M1 close-price series look more like a Weibull shape parameter of 1.5, more like a Poisson process of 5-pt jumps, or more like an exponential of the average price — and routes from there.

The probability layer is built in CalculateProbabilities(). On every OnTick, after UpdateData() loads lookback_period × 2 bars (default lookback_period = 1, so two bars each) from M1, M15, and H1, the EA computes three independent distribution fits over the M1 close series. The Weibull fit uses fixed shape 1.5 and scale 5.0 via MathProbabilityDensityWeibull(). The Poisson input is built in CalculatePoissonInput() by counting cumulative steps of 5-pt price changes in the M1 series, then feeding the running event counts through MathCumulativeDistributionPoisson() with lambda equal to the average event count (defaulted to 5.0 if it falls outside the (0, 1,000,000] sanity band). The Exponential fit uses MathCumulativeDistributionExponential() with mu equal to the M1 mean price, defaulted to 1.0 if non-positive. If any fit returns an empty array, every series falls back to a flat 0.5 — meaning the EA degrades gracefully but does not refuse to trade.

The classification is asymmetric, and that is worth knowing. IsBreakout() returns true when Poisson probability exceeds the breakout_threshold (default 0.4) AND Weibull probability exceeds the mean_reversion_threshold (default 0.4) AND the M1 close trend is up AND the M15 close trend is up AND H1 range (max − min) exceeds 10 points. When this fires, the EA opens a BUY only. IsMeanReversion() returns true when Weibull probability drops below 0.4 AND the absolute H1 trend is below 20 points. When this fires, the EA opens a SELL only. There is no path that fires a BUY from the reversion check or a SELL from the breakout check, and the only neutral exit is the protective SL/TP. Traders who want symmetric long/short exposure will need a different EA; this one is designed to be net-short on range sessions and net-long on trend sessions.

Stop placement is fully ATR-driven rather than fixed pips. GetATR() calls iATR(14) on the current chart symbol and timeframe, with a 10×point fallback if the handle is invalid. ExecuteTrade() then sets stop_distance = max(ATR × 1.5, 20 points) and target_distance = max(ATR × 2.0, 40 points), giving roughly 1:1.33 R:R at the floor and 1:1.33 R:R with wider stops when ATR dominates. m_validator.ValidateStopLoss() and m_validator.ValidateTakeProfit() add another 20% safety buffer on top of the broker's SYMBOL_TRADE_STOPS_LEVEL (with a 5-point floor when the broker reports zero), so the final SL/TP are pushed away from the entry by at least stopsLevel × point × 1.2. A BUY with a desired_sl of zero is rejected outright; the EA then submits the order with no SL — the CTradeValidator returns 0.0 for invalid SL/TP values rather than aborting the trade.

The CTradeValidator class is the most elaborate piece of infrastructure in the EX09 family. It encapsulates 16 helper functions: LoadSymbolInfo, Init, Refresh, CheckHistory, NormalizeVolume, ValidateVolume, CheckMarginForVolume, ValidateStopLoss, ValidateTakeProfit, ExecuteSafetyTrade, and the price-accessor getters. CheckMarginForVolume() requires 15% more free margin than OrderCalcMargin() reports, which is more conservative than the typical 5–10% buffer. ValidateVolume() walks the lot down by 25% increments when the requested size fails the margin check, stopping at SYMBOL_VOLUME_MIN. GetMarginDecimalDigits() returns 3 decimal places for symbols containing XAU, GOLD, XAG, or SILVER, and 2 elsewhere — this is how the EA shows lot sizes like 0.010 / 0.020 for gold rather than rounding them to whole pips.

Position sizing is delegated to the validator. CalculateLotSize() returns m_validator.ValidateVolume(BUY, 0.01) for normal operation, which means the practical minimum is 0.01 lot on FX and 0.01 lot on gold when the broker allows it. For XAUUSD with a sub-500 balance or in a detected validation run, the function short-circuits to m_validator.GetMinLot() — there is no percentage-of-equity lot math in this EA. The risk_per_trade = 1.0 input is therefore nominal and not directly used in CalculateLotSize; it appears to feed only the Risk label in the header comment. The InpCapAmount / InpCapFloor pair (defaults 0.0 / 50.0) implements a hard capital ceiling: EffectiveCapital_EX09019() returns min(InpCapAmount, equity) and the daily loss limit in IsSafeToTrade_EX09019() is computed against that effective capital.

The nine-gate pre-trade safety stack is standard EX09 hardening. IsSafeToTrade_EX09019() blocks trades when: the cap is reached (equity >= InpCapAmount when the cap is on), the floor is breached (equity < InpCapFloor), equity has fallen below 95% of g_initialBalance, realized today is below −3% × EffectiveCapital, three consecutive losses are still inside the 30-minute cooldown window, the 50-trade daily ceiling is hit, the market is closed (Friday 21:00 GMT close through Sunday 22:00 GMT reopen), the active session is empty (London 7–16, NY 12–21, with InpAvoidAsia=true blocking 0–7 when neither London nor NY is open), or — if InpNewsFilter is on — within ±15 minutes of the London or NY session start hour. Realized P&L is fed by OnTradeTransaction() which appends profit + swap + commission on every DEAL_ENTRY_OUT deal whose magic matches Magic_Number.

Three tester-specific behaviors are worth highlighting. First, DetectValidationEnvironment() flips m_is_validation_run = true when balance <= 500 OR the symbol/timeframe combo matches one of four hardcoded test pairs (EURUSD H1, XAUUSD D1, GBPUSD M30, EURUSD M1). When this flag is on, OnTick() short-circuits to a safety trade whenever no signal fires and no position is open, guaranteeing that the strategy tester never returns a zero-trade backtest. Second, CheckForSafetyTrade() calls m_validator.ExecuteSafetyTrade() (a minimum-lot market order with magic 999999) every safety_trade_interval seconds, gated by enable_safety_trade. Third, SaveArraysToFile() and LoadArraysFromFile() persist the seven CArrayDoubleEx buffers to BreakRevert_Data.dat every 1000 ticks on a live account (suppressed in the tester), so the probability arrays survive terminal restarts.

The send wrapper is conservative. TryBuy_EX09019 and TrySell_EX09019 both run OrderCalcMargin first, then submit via CTrade::Buy / CTrade::Sell with SetDeviationInPoints(10), retrying up to 3 times on TRADE_RETCODE_REQUOTE, TRADE_RETCODE_TIMEOUT, TRADE_RETCODE_PRICE_OFF, or TRADE_RETCODE_PRICE_CHANGED with a 200 ms Sleep between attempts. Any other retcode breaks the loop. The m_trade.SetExpertMagicNumber(Magic_Number) call on init means every position opened by the EA carries magic 22209019, so the 90-day deal-history filter in OnTradeTransaction() only counts exits with that magic.

There is one practical bug in the input block that traders should know about. The TF_M1/TF_M15/TF_H1 inputs accept integers 1 through 7, but the defaults are 1, 6, and 8 respectively. Since 8 falls outside the 1–7 range, MapTimeframeInt() returns PERIOD_H1 for both g_TF_M15 and g_TF_H1, meaning the M1 series is genuinely M1, but the M15 and H1 series are both pulled from H1. Anyone running this EA on a chart other than H1 should set TF_M15 = 3 and TF_H1 = 5 to get the multi-timeframe behavior the comment promises. With the bug, the EA collapses to M1+H1+H1 — still a multi-TF read, but not what the description implies.

For backtest expectations, the right reference period is 2019–2024 on XAUUSD M5 with the default inputs. Expect a modest trade count (50–120 trades per year given the 10-second trade delay and 0.4 probability threshold) and a high win rate on the mean-reversion side during Asia/London rollover windows, where H1 trends stay below 20 points. The breakout path needs genuine M15 momentum to fire, so it will sit idle during long consolidations. Optimize breakout_threshold and mean_reversion_threshold first; they govern almost every trade. OnTester ranks candidates by netProfit × profitFactor / (1 + relativeDD%) with a 30-trade minimum, which heavily penalizes over-fitting to high net-profit but high-drawdown configurations.