Pipsgrowth EX05003 Hedge

Pipsgrowth EX05003 is an asymmetric dual-engine USDJPY hedger built around a single observation: yen pairs grind upward for weeks and then collapse in hours when the Bank of Japan talks tough or the U.S. CPI print surprises. The EA captures both sides of that pattern with two completely separate signal stacks that share an account but use independent magic numbers (InpMagicLong = 22205003 for the long engine, InpMagicShort = 22215003 for the short engine). On the long side it is a patient SMC follower that only buys when an H1 swing structure has just turned bullish, the close retests a freshly created bullish order block, and the D1 simple moving average (MA_D1_PERIOD = 50) is sloping up beneath the bid. On the short side it is a volatility-event hunter that only sells when the bid is above the user-defined intervention zone (InpInterventionLevel, default 150.0 — a price that corresponds to a typical USDJPY level the BoJ has historically defended) and the previous daily low (g_prevDayLow, refreshed each new bar from iLow(_Symbol, PERIOD_D1, 1)) has just been broken. The two engines never blend into a single basket: each is sized, stopped, trailed, and partial-taken by its own function set (ManageLongPositions, ManageShortPositions) and the two g_blockReason strings (e.g. CAP_BELOW_FLOOR, DAILY_LOSS, MARGIN) are tracked independently so you can see at a glance which leg tripped a guard.

The regime layer that gates both engines is a seven-state classifier built from three indicators on the chart timeframe — ADX(14) via g_hADX, a 20-period/2.0-deviation Bollinger Band via g_hBB, and a 14-period ATR via g_hATR — combined with an H1 ADX mirror (g_hADX_H1) and the daily 50-MA. The UpdateRegime() function reads ADX bar 1, computes an ATR-to-close percent and a Bollinger band width proxy, then runs an ATR-percentile rank over a 100-bar lookback (ATRPCT_LOOKBACK) before resolving the state. The long engine refuses to seed in REG_CHOPPY, while the short engine additionally refuses to seed in REG_COMPRESS and REG_RANGE — shorts need volatility to confirm a PDL break is a real event, not a thin-session poke. STRONG_TREND (ADX ≥ 30) and WEAK_TREND (ADX ≥ 18) add 20 confidence points to the long signal, and BREAKOUT / EXPAND add 20 to the short signal. CHOPPY is a hard refusal for both sides, which is what keeps the EA from getting chopped up during Tokyo's pre-open drift.

Entries are driven by LongSignal() and ShortSignal(). LongSignal returns +1 only when bid is above the D1 MA(50), the H1 swing trend (g_swingTrend) is BULLISH, the last closed bar on the chart timeframe was a bullish candle (close > open), and the low of that bar touched an active bullish order block (OBTouch(BULLISH, low1)). Order blocks themselves are built by CreateOrderBlock(bias, refBar), which walks back up to MAX_OB_LOOKBACK = 10 H1 bars from the swing pivot, picks the first candle whose body runs in the direction of the bias (a bearish candle for a bullish OB, a bullish candle for a bearish OB), and records its high and low. There are at most MAX_OB_COUNT = 5 OBs in the ring buffer; the oldest is shifted out when a sixth is added. CheckOBMitigation() deactivates a bullish OB the moment the H1 low closes below its bottom (and a bearish OB the moment the H1 high closes above its top), and ExpireOldOBs() retires anything older than MAX_OB_AGE_BARS = 100 H1 bars. The combination of mitigation + expiry + retest is what makes the long engine selective: the same OB can only trigger once, and only inside its 100-bar window.

ShortSignal returns -1 only when bid ≥ InpInterventionLevel, g_prevDayLow is populated, and the last closed chart bar has either its close or low below g_prevDayLow. Confidence starts at 65 and climbs another 20 in BREAKOUT or EXPAND regimes, 10 if ATR-percent is above 0.5 (intervention-grade volatility), and 5 if the H1 swing structure is already bearish. That asymmetry is intentional: shorts into a PDL break during intervention-grade volatility are the only short signal worth taking, and the regime and confidence stacking makes the trade conditional on actual expansion. Once both ConfirmLong / ConfirmShort pass — ConfirmLong requires a non-CHOPPY regime, the HTF gate, and a valid R:R (MIN_RR_RATIO = 1.0 enforced via InpRR_Long / InpRR_Short); ConfirmShort requires non-COMPRESS, non-RANGE, non-CHOPPY, and a valid R:R — the EA sizes the order in CalcLot(slDistance, effectiveCap) against ATR-based stop distance, applying a hard ceiling at InpMaxLot (default 5.0 lots) and normalizing to the broker's volume step. Margin is pre-checked via OrderCalcMargin and refused if the new order would eat more than 80 percent of the free margin.

Management is where the two engines really diverge. Long stops sit 2.5 × ATR(14) below entry (InpSLATR_Long) — deliberately wide because USDJPY spikes 30-50 pips on every BoJ rumor and a tight stop would die on noise. Short stops sit 1.5 × ATR(14) above entry (InpSLATR_Short) — deliberately tight because once a PDL breaks during intervention, the move is fast and the EA wants to be wrong small rather than wrong by a hundred pips. Long R:R is 1.5 (InpRR_Long), short R:R is 3.0 (InpRR_Short), which means the short side is a 1:3 setup on a 1.5-ATR stop, the long side a 1:1.5 on a 2.5-ATR stop. Once profit exceeds TRAIL_ACTIVATE_ATR = 1.0 × ATR the stop is ratcheted to entry plus BE_BUFFER_POINTS = 10 points (break-even) one-shot, and a 50 percent partial is taken at TP1 (PARTIAL_TP_FRACTION = 0.5) provided the residual volume is at least 1.5 × the broker's minimum lot. After break-even, the stop is trailed at InpTrailATR = 1.5 × ATR(14) using the H1 ATR for longs and the chart-timeframe ATR for shorts, and any time the swing trend flips against the position the EA closes it via TryClose_EX05003 (3 retries at 200 ms each on REQUOTE / TIMEOUT / PRICE_OFF / PRICE_CHANGED). Friday 20:00 server time triggers CloseAllByMagic for both magics if InpCloseFriday is on, so no trade survives the weekend gap.

The hardening layer is the same safety stack the rest of the EX05 hedge family runs. Effective capital is min(InpCapAmount, equity) when a cap is configured, plain equity otherwise, and EffectiveCapitalNet() subtracts the day's realized PnL across both magics so the lot calculator sees the actual headroom. Daily loss limit is InpDailyLossLimitPct = 3 percent of effective capital, and the weekly gate is twice that. InpMaxConsecLosses = 3 losses arms a 30-minute cooldown (g_cooldownUntil), which is enforced in IsSafeToTrade along with the per-day trade cap (InpMaxTradesPerDay = 50), the capital floor (InpCapFloor = $50 by default — any equity below this blocks new entries), the 95 percent minimum equity floor (InpMinEquityPercent), and the IsMarketOpen / InActiveSession / IsNewsTime gates. The session filter is GMT-aware via DetectGMTOffset, which walks back up to 100 H1 bars looking for the first weekend gap > 2 hours and infers the server's offset from there. With InpAvoidAsia left at false (the default — this is a Tokyo EA, not an anti-Tokyo EA) it accepts Tokyo 0-9 broker time, plus London 7-16 GMT and New York 12-21 GMT. Order entry is 3 retries on REQUOTE / PRICE_OFF / PRICE_CHANGED, 100 ms apart; SLs are clamped to the broker's stop level plus the BE buffer via ClampSLForBuy / ClampSLForSell, and the OnTester custom criterion is (net × profit factor) / (1 + max DD) with a 30-trade minimum. The dashboard Comment() prints regime, ADX, ATR%, BB width, and the current block reason on every tick.

In backtest, expect the long engine to take the majority of trades during sustained D1 uptrends in USDJPY, with most of the PnL coming from the 50 percent partial + trailing stop on those longer swings. The short engine will be quieter — Tokyo intervention breaks are rare, and the regime + confidence + intervention-level gates mean a typical year might produce only a handful of short seeds, each with a tight 1.5-ATR stop and a 3R target. Drawdown profiles of the two engines are independent, which is the point: a short that catches a BoJ surprise can print more in one trade than the long engine prints in a month, and the long engine's wider 2.5-ATR stop is the cost of staying in the trade long enough to give the trailing stop room to work. The two magic numbers make it straightforward to run the long engine and the short engine on different charts of the same account, or to disable one side entirely (InpEnableLongs / InpEnableShorts) when the user wants to convert the EA into a pure SMC long-only or a pure intervention-short specialist.