Princeec13/mql5
0
0
Bifurcar 1
Código Incidencias Pull requests 1 Proyectos Lanzamientos Paquetes Wiki Actividad Acciones
mql5/Experts/Advisors/DualEA/docs/Phase3.md
Princeec13 a5a81e4cfc new files added
2026-04-14 19:36:52 -04:00

8 KiB
Original Enlace permanente Blame Histórico

Phase 3: LiveEA Core Loop & Documentation Lockdown

Overview

Phase 3 focuses on completing the LiveEA core trading loop with policy gating, insights integration, risk management, and comprehensive documentation alignment. This phase bridges the gap between PaperEA (data collection) and LiveEA (production trading).

Current Status

✅ Completed

  • LiveEA.mq5 (1832+ lines): Core trading loop with policy loading and gating
  • LiveEA_StrategyBridge.mqh: Strategy bridge for 21+ IStrategy implementations
  • InsightsLoader.mqh: DRY insights parsing from insights.json
  • PolicyEngine.mqh: ML policy loading with confidence thresholds
  • Telemetry.mqh: Real-time telemetry with standardized event schema
  • PositionManager.mqh: Correlation-aware sizing and dynamic risk caps
  • IncrementalInsightEngine.mqh: Real-time O(1) statistics engine

🔄 In Progress

  • LiveEA hardening with stricter risk gates
  • Timer-scan parity with PaperEA (Phase 6 alignment)
  • Shadow mode for candidate decision logging

Architecture

LiveEA Execution Pipeline

OnTick() / OnTimer()
    ↓
ScanStrategies() → Per-strategy signal generation
    ↓
InsightGating() → Check performance metrics per slice
    ↓
PolicyGating() → ML confidence scoring with scaling
    ↓
RiskGating() → Spread, session, drawdown checks
    ↓
CorrelationCheck() → Portfolio exposure validation
    ↓
ExecuteOrder() → Trade placement with audit trail
    ↓
LogToKnowledgeBase() → Record for feedback loop

Key Components

1. Policy Integration (PolicyEngine.mqh)

  • Loads policy.json from Common Files on init
  • ApplyPolicyScaling(): SL/TP/trailing scaling per slice
  • Fallback modes: fallback_no_policy, fallback_policy_miss
  • Demo-only restriction configurable via FallbackDemoOnly

2. Insights Gating (InsightsLoader.mqh)

  • Real-time loading of insights.json performance metrics
  • Per-slice validation: win rate, expectancy, drawdown
  • Auto-reload with InsightsAutoReload polling mechanism
  • Freshness tracking via file modification time

3. Risk Management (CPositionManager.mqh)

  • Correlation-aware position sizing
  • Dynamic risk caps with daily drawdown limits
  • Kelly Criterion position sizing (simplified)
  • Volatility-based exit triggers

4. Telemetry (Telemetry.mqh / TelemetryStandard.mqh)

  • Event-driven logging with severity levels
  • Per-slice gating counters and latency metrics
  • Cross-EA compatibility with PaperEA
  • Shadow logging under NoConstraintsMode

Input Parameters

Core Settings

Parameter Type Default Description
MagicNumber long 123456 Trade identification
NoConstraintsMode bool false Shadow-gate only, don't block
Verbosity int 1 0=errors, 1=warn, 2=info, 3=debug

Risk Management

Parameter Type Default Description
SpreadMaxPoints double 0.0 0=disabled
MaxDailyLossPct double 0.0 Daily loss limit
MaxDrawdownPct double 0.0 Max drawdown %
ConsecutiveLossLimit int 0 Max consecutive losses

Insights Gating

Parameter Type Default Description
UseInsightsGating bool true Enable insights validation
GateMinTrades int 0 Min trades for slice validity
GateMinWinRate double 0.00 Min win rate threshold
GateMinExpectancyR double -10.0 Min expectancy in R

Policy Gating

Parameter Type Default Description
UsePolicyGating bool true Enable ML policy gating
DefaultPolicyFallback bool true Allow neutral trading on missing slice
FallbackDemoOnly bool true Restrict fallback to demo accounts
FallbackWhenNoPolicy bool true Allow fallback when no policy file

Exploration Mode

Parameter Type Default Description
ExploreOnNoSlice bool true Allow limited trades for new slices
ExploreMaxPerSlice int 100 Max exploration trades per slice
ExploreMaxPerSlicePerDay int 100 Daily exploration cap

PositionManager Integration

Parameter Type Default Description
UsePositionManager bool false Enable PositionManager integration
PM_EnableCorrelationSizing bool true Apply correlation dampening
PM_CorrMinMult double 0.5 Min multiplier after dampening
PM_CorrMinLots double 0.01 Absolute minimum lots floor

Feedback Loop

The LiveEA → PaperEA feedback loop enables continuous improvement:

  1. LiveEA executes trades and logs to knowledge_base.csv
  2. ML Pipeline merges PaperEA + LiveEA data for training
  3. Trainer generates updated policy.json with new patterns
  4. PaperEA validates policy changes in simulation
  5. Promotion Gate advances validated strategies to LiveEA
LiveEA ──trades──→ knowledge_base.csv
                          ↓
PaperEA ──features──┴──→ ML Pipeline
                          ↓
                   policy.json (retrained)
                          ↓
         ┌────────────────┴────────────────┐
         ↓                                 ↓
    LiveEA (hot-reload)              PaperEA (validation)

3×3 Red-Team Execution Plan

Phase 1: Codebase & Requirements Deep Dive

  • Cycle 1: Structural Audit for LiveEA

    • Inventory gating paths: strategy → selector → insights → policy → execution
    • Trace file I/O: CheckPolicyReload(), CheckInsightsReload(), staleness checks
    • Identify insertion points: circuit breakers, session filters, PositionManager

  • Cycle 2: Requirements Mapping

    • Spec one-pagers per feature with inputs, data flow, telemetry, failure/rollback

  • Cycle 3: Risk & Gaps Closure

    • Threat model: file races, stale artifacts, partial reloads
    • Define LiveEA fail-safe defaults

Phase 2: Design, Build, Validate

  • Cycle 1: Task decomposition for LiveEA hardening
  • Cycle 2: Implement fixes; peer review and adversarial edge-case tests
  • Cycle 3: QA + fuzzing of IO/races; enforce CI gates

Phase 3: Polish, Lockdown, Final Scoring

  • Cycle 1: Docs & architecture alignment (README + this file)
  • Cycle 2: Code hygiene & consistency; refactors; dead-code purge
  • Cycle 3: Final scoring + roadmap to Phase 4 (Risk & Safety Systems)

Telemetry Events

Key Event Types

  • live_init: EA initialization with policy/insights status
  • policy_loaded: Policy file successfully loaded
  • policy_fallback_*: Fallback mode activation
  • insight_gating: Per-slice insights validation result
  • risk4_*: Risk gate triggers (spread, session, drawdown)
  • p5_*: Phase 5 selector events (cooldown, auto-disable, re-enable)
  • pm_corr_sizing: PositionManager correlation sizing adjustment
  • order_executed: Trade execution with full context
  • order_blocked: Trade blocked with gating reason

Event Schema

{
  "timestamp": "2026-01-15T14:30:00Z",
  "event": "insight_gating",
  "symbol": "EURUSD",
  "timeframe": 60,
  "strategy": "ADXStrategy",
  "allowed": true,
  "metrics": {
    "trades": 150,
    "win_rate": 0.58,
    "expectancy_r": 0.35,
    "max_drawdown_r": 2.1
  }
}

Next Steps

  1. Complete Phase 3 Cycle 3: Final scoring and documentation alignment
  2. Begin Phase 4: Risk & Safety Systems hardening
  3. Phase 6 Alignment: Timer-scan parity with PaperEA
  4. CI Integration: Wire ValidateInsights.mq5 into automated testing

Related Documentation