Pipsgrowth EX09010 MeanReversion

The Pipsgrowth EX09010 "Thanos Snap" sits squarely in the Mean Reversion family, but its personality comes from a single signature feature: a recovery grid that snaps in below the original entry when the first trade goes against it. Despite the dramatic name, the grid is hard-capped, multiplier-bounded, and disabled by default — turn it on, and the EA behaves like a controlled martingale; leave it off, and it runs as a clean dual-oscillator reversal system. That binary is the most consequential choice a user will make.

Signal construction is the second distinguishing trait. EX09010 reads both RSI(14) and CCI(14) on the last closed M5 bar and only fires when both oscillators agree at the threshold — RSI below 30 paired with CCI below -100 for a long entry, RSI above 70 paired with CCI above +100 for a short entry. A confidence score is logged (50 + 2× the depth beyond the threshold, clamped 20-100) but is informational, not gating. The signal will not fire off either oscillator alone; that AND-logic is what filters the bulk of the false reversals on trending days and makes the dual-threshold design worth more than the sum of its parts. Other Mean Reversion EAs in this family mix RSI with an EMA, or CCI with a Stochastic cycle swing, or use a candlestick-pattern gate — this one is specifically dual-oscillator AND-confirm.

The trade is then handed to a confirmation layer that reads the H1 EMA(50) and requires price to be extended at least 0.25×ATR(14) from it. For a long reversal, the EA wants the bid sitting more than a quarter-ATR below the H1 line; for a short reversal, more than a quarter-ATR above. This is the "pullback context" check, and it is what separates a genuine overshoot from a mid-range noise spike. Without it, a Mean Reversion EA will keep trying to fade healthy intraday pullbacks inside an H1 trend; with it, EX09010 only triggers when the M5 has stretched meaningfully away from where the higher timeframe thinks value lives.

Regime is handled by a 7-state classifier that runs on ADX(14), Bollinger Bands (20, 2.0) width ratio, and ATR(14) percentile. The states are StrongTrend, Expand, WeakTrend, Range, Compress, Choppy, and an unknown fallback. The NoTrade gate explicitly blocks StrongTrend and Expand — both flags the EA reads as "do not fade a trend in motion." WeakTrend, Range, and Compress are the productive states for the dual-oscillator reversal. Choppy is permitted for the entry itself but is what force-closes an open position if the regime flips after entry. This is a deliberate separation: a Choppy open is fine to enter, but a Choppy post-entry regime is treated as a deteriorating condition that should not be held.

Risk is sized at 0.5% of effective capital per trade, with the lot calculation reading both tick_value and tick_size off the symbol and dividing the risk budget by the per-lot loss at the configured SL distance. SL is set at ATR(14)×2.0 (InpSignalATRMult=2) and floored at 1.2× the broker's stop level so it never sits inside the freeze zone. TP sits at 2.0R from entry (InpTP_R=2.0), which gives an asymmetric R:R that the partial-close logic further exploits. The capital allocation cap layer applies a hard ceiling: InpCapitalCapAmount, when set above zero, replaces real account equity with min(cap, equity) for all percentage-based math. A floor of 50 USD (InpCapitalCapFloor) blocks new entries entirely when equity falls below it. The realized PnL windows (3% daily, 6% weekly) are tracked per-magic and reset at the day and week boundaries.

The position management layer stacks five exit paths. The +1R break-even fires first: when unrealized profit reaches one unit of risk, the stop is moved to entry+0.2R (so the trade is essentially free unless the market round-trips to entry). The 50% partial close at +1R follows, but only if the remaining volume clears 1.5× the broker minimum lot — small accounts and small-size grid layers can skip the partial. The ATR(14)×2.5 trailing stop activates at +0.8R and only ratchets forward, never back. The 80-bar time stop on M5 (= 6 hours 40 minutes of bars) is the hard ceiling for any single trade. The fifth path is the regime-flip: if the classifier reports Choppy at any manage tick, the open position is force-closed immediately.

The Thanos Snap recovery grid is the section most users will read first. When InpEnableRecovery is true (false by default), ManageRecovery() watches the most recently opened position of the EA's magic. If price moves against it by InpRecoveryStep=300 points and the count is below both InpMaxConcurrent=5 and the hard-coded RECOVERY_MAX_LEVELS=5, the EA opens a same-direction layer sized at lastLot × min(InpRecoveryMult=1.4, RECOVERY_MAX_MULT=1.5). The same NoTrade gates apply — no recovery during weekends, no recovery when spread exceeds InpMaxSpread=250 points, no recovery during Choppy regimes. The hard caps on count and multiplier are deliberate: the EA explicitly refuses to convert into a runaway martingale even if the user pushes the inputs high.

A few features this EA does not include are worth flagging. There is no HTF bias (only an HTF extension check), no equity-floor stop, no news feed integration, no time-of-day session filter, and no pyramiding beyond the optional recovery grid. OnTester is implemented as (net × profit_factor) / (1 + equity_dd_pct) with a 30-trade minimum, rewarding both profitability and low drawdown simultaneously. SetTypeFillingBySymbol is auto-called in OnInit so the EA negotiates FOK versus IOC versus RETURN based on what the broker offers. All trade-send retries are 3-iteration loops with 200ms sleep on REQUOTE/TIMEOUT/PRICE_OFF/PRICE_CHANGED; modify retries use 100ms and only on REQUOTE/TIMEOUT. TryClosePartial exists but is called only by the partial-TP path inside ManagePositions — no other partial-close triggers are wired.

The most important footgun is InpDryRun=true, which is the default. With dry-run on, the EA logs every would-be entry but never actually opens one — useful for paper testing, brutal for anyone who downloads the EA and attaches it to a live chart without flipping the toggle. Pair that with InpKillSwitch (defaults to false) and you have two booleans the operator must check before expecting the EA to actually trade. The dry-run is also the reason recovery logic short-circuits at the top of ManageRecovery(): stacking grid levels in a paper test would generate noise without information.

What to expect on a backtest: the dual-oscillator signal is sparse but clean, so the EA will tend to take 2-5 trades per day on XAUUSD M5 under normal conditions, with most entries clustering on Range and Compress regime days. The recovery grid, when enabled, multiplies the count significantly because every losing first entry becomes the trigger for 1-4 same-direction layers. Expect the equity curve to stair-step when the grid is on — fast recovery on shallow reversals, drawn-out recovery on the deep runs that test the 5-level cap. Without the grid, the curve looks closer to a classic mean-reversion equity: many small wins, occasional full-stop losses, and a steady return profile that depends almost entirely on regime selection rather than signal quality.