Random Forest Decision EA
Eight expert modules vote. Confidence threshold gates every trade. Ensemble decision-making for MT5.
A deterministic pseudo-Random-Forest ensemble engine for MetaTrader 5. Eight independent expert modules — Trend, Momentum, Volume, Volatility, Structure, HTF, ADX, and Candle — each cast a directional vote. A weighted confidence score gates every entry, with four model profiles (Balanced, Trend-Heavy, Momentum-Heavy, Conservative) to match your style. Feature-randomization seed lets you run multiple instances with slightly varied emphasis — true ensemble diversification on one account.
What makes it different.
- Eight independent expert modules — Trend (3-stack EMA), Momentum (RSI+MACD+Stoch), Volume (OBV+pressure), Volatility (Bollinger+squeeze), Structure (adaptive pivots), HTF (multi-timeframe EMA), ADX (regime strength), and Candle (body-ratio + engulfing)
- Weighted confidence score — configurable threshold (default 0.62) and min-votes gate (default 5-of-8) ensure trades only fire on strong multi-factor agreement
- Four model profiles — Balanced (equal weights), Trend-Heavy (Trend + HTF x2, ADX x1.5), Momentum-Heavy (Momentum + Candle x2, Volume x1.5), Conservative (any neutral vote kills the signal)
- Feature randomization seed — deterministic bagging-style variation lets you run 3-5 instances with different seeds on the same chart for genuine ensemble diversification
- HTF confirmation gate — optional filter requires higher-timeframe EMA alignment to agree with the overall signal direction
- ADX regime filter — below-threshold ADX readings return neutral votes from the ADX expert, protecting against false signals in low-trending markets
- Full risk stack — fixed-lot or percent-risk sizing, ATR stop-loss and take-profit, ATR trailing stop, break-even logic, optional partial-close at 1R
- Session + frequency filters — hour-of-day trading window, per-weekday toggles, daily trade cap, cooldown bars after each entry, daily loss kill-switch
- On-chart decision dashboard — live vote tally, confidence score, active profile, regime state, spread, and drawdown telemetry
- Preset library — Balanced Intraday (EURUSD M15), Trend-Heavy Swing (H1), Conservative Prop (M30), Gold Specialist (XAUUSD H1), Index Runner (NAS100 M15)
Simulated on $10k starting equity at the stated CAGR. Past performance does not guarantee future results. See risk disclosure.
Every input, documented.
| Parameter | Default | What it controls |
|---|---|---|
| InpModelProfile | BALANCED | Weighting scheme for the 8 modules. BALANCED (equal), TREND_HEAVY, MOMENTUM_HEAVY, CONSERVATIVE (neutral-vote veto). |
| InpMinVotes | 5 | Minimum non-neutral votes (out of 8) required to trigger an entry. Range 1-8; 5 is the tuned default. |
| InpConfidenceThresh | 0.62 | Weighted confidence score required to trigger a trade. 0.50-1.00. Higher = fewer, higher-quality signals. |
| InpFeatureSeed | 42 | Deterministic seed for feature-subset variation. Different values produce slightly different emphasis patterns — analogous to individual trees in a Random Forest. |
| InpUseHTFConfirm | true | Requires the HTF Expert to return a non-neutral vote aligned with the overall signal direction. |
| InpUseADXFilter | true | Activates the ADX regime filter. Low-ADX bars neutralize the ADX vote rather than permitting range entries. |
| InpAdxMinLevel | 20.0 | ADX value below which the ADX Expert returns a neutral (0) vote. |
| InpRiskMode | RISK_PERCENT_SL | Fixed lot or percent-risk-on-SL-distance sizing. |
| InpRiskPercent | 1.0 | Equity % risked per trade when using percent-risk mode. |
| InpAtrPeriod | 14 | ATR period used for all SL/TP/trailing computations. |
| InpAtrSlMult | 1.5 | ATR multiplier for the initial stop-loss distance. |
| InpAtrTpMult | 2.5 | ATR multiplier for the take-profit distance. |
| InpTrailingStop | true | Enables ATR-based trailing stop. |
| InpBreakEven | true | Moves SL to entry when price reaches 1R profit. |
| InpMaxDailyTrades | 5 | Maximum new trades per calendar day. 0 disables the cap. |
| InpCooldownBars | 3 | Bars to wait after a trade opens before a new entry can fire — prevents overtrading during volatile bursts. |
| InpMaxDailyLossPct | 3.0 | Daily loss kill-switch. Halts new entries for the session once the loss exceeds this percentage. |
| InpSpreadMax | 30 | Maximum spread (points) permitted at entry. Protects against news-window blowouts. |
| InpMagic | 20250001 | Magic number. Use a unique value per instance to run multiple seeds concurrently. |
Full parameter reference in the PDF user guide included with your license.
Frequently asked.
Is this a real Random Forest or a deterministic approximation?
A deterministic approximation. Eight hand-engineered expert modules cast votes in place of trained decision trees, and the feature-randomization seed mimics bagging's feature-subset variation. This keeps the EA fully reproducible — no model file to load, no training pipeline — while preserving the core ensemble benefit of noise cancellation across independent experts.
How does this differ from kNN Machine Learning EA?
kNN ML classifies the next bar from similarity to historical patterns (one model). Random Forest Decision aggregates eight independent views — one of which is price action (candle structure), one is volatility regime, one is market structure. They solve different problems: kNN is sample-driven; Random Forest is consensus-driven.
Can I run multiple instances with different seeds?
Yes — that's the intended power-user setup. Launch 3-5 instances on the same chart with seeds 42, 137, 256, 512, 1024 and unique magic numbers. Each instance will have slightly different emphasis per expert module. Run them at 25% of your usual position size each for genuine ensemble diversification.
Which model profile should I start with?
Balanced for forex majors on M15-H1. Trend-Heavy for H4+ swing accounts. Momentum-Heavy for indices and gold (where momentum leadership matters). Conservative when you're running a prop challenge and want the tightest filtering.
Is it prop-firm safe?
Yes. The Conservative profile combined with a 0.70 confidence threshold and 6-of-8 minimum votes produces roughly 30-50% fewer trades with a significantly lower DD profile. Under the default settings, backtest DD is 8.9% — inside the 10% FTMO cap — but tighten to Conservative + 0.70 for extra margin.
What symbols work best?
Forex majors (EURUSD, GBPUSD, USDJPY), XAUUSD, and major indices (NAS100, US30, GER40). The EA is agnostic about instrument — what matters is that the spread is reasonable and the HTF EMA structure is legible. Avoid exotics where the volume expert's OBV calculation is unreliable.
Does it repaint?
No. All expert votes are computed on closed bars; no module reads in-progress bar data. The confidence score is latched at the bar close that triggers the entry.
Ready to run Random Forest Decision EA?
Lifetime license. All presets included. Direct developer support.