Pavan Gold Regime Engine
- エキスパート
- バージョン: 1.10
- アクティベーション: 5
Advantages (what genuinely sets it apart)
- Adapts instead of assuming — most EAs are one strategy applied blindly. This one classifies the market regime on every closed bar and re-weights four different strategies to fit it. It has a "stay out" decision built in — no signal, no trade.
- Machine-trained on real tick data — parameters weren't hand-guessed; they came from walk-forward optimization over 2.5 years of real XAUUSD ticks, validated on a 6-month holdout the optimizer never saw (+30% / PF 1.15 on that unseen period).
- Capital preservation is enforced, not promised — stop loss and take profit attached to every order at open; the EA cannot place a naked position. No martingale, no grid, no averaging down, no recovery ladders — the account-killer patterns are structurally impossible.
- Risk is defined in percent, not lots — position size is computed per-trade from the SL distance, so a stop-out costs the configured % of equity regardless of volatility.
- Knows when to stop each day — books the day after hitting the profit target, halts after the daily loss floor, caps trades per day.
- Transparent — every entry logs its confidence, agreeing strategies, regime, and R:R to the Experts journal.
Features
Analysis engine (every closed M15 bar)
- Regime detector: trend-up / trend-down / range / quiet (volatility compression) / volatile chop — from ADX, EMA-stack + slope, and ATR percentile
- Voter 1 — Trend-following: H1 alignment, EMA20/50/200 stack, ADX strength, MACD momentum + acceleration, session-VWAP side, pullback quality, RSI overextension veto
- Voter 2 — Mean reversion: Bollinger %B stretch, RSI exhaustion, stochastic confirmation, distance-to-mean check
- Voter 3 — Breakout: 20-bar range break in ATR units, tick-volume expansion, Bollinger squeeze release, candle-body decisiveness (fakeout filter)
- Voter 4 — Smart Money Concepts: liquidity sweep + reclaim, BOS, CHOCH, order blocks, fair value gaps, premium/discount zone location
- Regime-weighted probability blend → single confidence score; conflict between voters lowers it; thin sessions penalize it
Trade management
- ATR-based SL and TP on every order · breakeven jump at +X R · ATR trailing stop after +1R · 50% partial profit banking at +X R · time-based exit after N bars
Risk & safety
- Fixed %-of-equity risk sizing with broker lot-step/min/max normalization · daily profit booking halt · daily loss floor · max trades/day · spread filter · one position at a time · magic-number isolation (won't touch other trades)
Input parameters
Signal gates
| Input | Default | Meaning |
|---|---|---|
| InpMinConfidence | 0.334 | Minimum blended confidence to trade (higher = fewer, stricter trades; range 0.3–0.7) |
| InpMinAgree | 2 | Voters that must agree (a single voter ≥0.8 also qualifies) |
| InpMinRR | 1.23 | Minimum risk:reward after SL/TP construction |
| InpSLmultATR | 2.39 | Stop distance = max(ATR × this, 0.5) |
| InpTPmultATR | 5.99 | Target distance = max(stop × MinRR, ATR × this) |
| InpAllowVolatile | false | Trade the high-volatility chop regime (safer off) |
| InpRequireLiquidSession | false | Hard-block entries outside London/NY hours |
Trade management
| Input | Default | Meaning |
|---|---|---|
| InpBreakevenAtR | 1.48 | Move SL to entry once trade is +X R (0 = off) |
| InpTrailATR | 2.5 | Trail SL at X×ATR behind price after +1R (0 = off) |
| InpPartialAtR | 1.2 | Close 50% of the position at +X R, let the rest run (0 = off) |
| InpMaxHoldBars | 48 | Time exit after N M15 bars (~12 h) |
Risk & limits
| Input | Default | Meaning |
|---|---|---|
| InpRiskPct | 1.0 | % of equity risked per trade (1–2 sane, 5+ aggressive, tested to destruction at 10) |
| InpDailyProfitPct | 10.0 | Booking halt: stop trading for the day at +X% (10% = $1,000 on $10k) |
| InpDailyLossPct | 3.0 | Daily loss floor: stop trading for the day at −X% |
| InpMaxTradesPerDay | 6 | Entry cap per calendar day |
| InpMaxSpreadPoints | 50 | Skip entries when spread exceeds this |
Technical
| Input | Default | Meaning |
|---|---|---|
| InpBrokerUTCOffset | 2 | Broker server time minus UTC (for session detection) |
| InpMagic | 20260708 | Order tag so the EA only manages its own trades |
Honest footnote for any listing: management defaults (partial at +1.2R, trail 2.5) are currently hand-set per your spec — the tester showed they trade profit for comfort versus plain targets (+48% vs +75% on the year). The ultra optimizer running right now is searching for the combination that keeps both; when its holdout verdict lands, these defaults get replaced with trained values. Recommended demo period before any live use: 4+ weeks.
