foxhunt/PAPER_TRADING_FIX_SUMMARY.md

Paper Trading Fix - Executive Summary

Agent 131 | Date: 2025-10-14 | Status: 🔴 CRITICAL BUG IDENTIFIED

---

Problem

3,000 predictions → 0 orders (0% conversion rate)

---

Root Cause

Missing paper trading executor service

The ML ensemble is generating predictions and logging them to ensemble_predictions table, but no code exists to:

1. Read predictions from database
2. Filter by confidence threshold
3. Create orders in orders table
4. Link predictions to orders

---

Evidence

Predictions: ✅ WORKING

SELECT COUNT(*) FROM ensemble_predictions;
-- 3000 predictions
-- Generated: 2025-10-14 15:06-16:06 UTC
-- Symbols: TEST_SYM only
-- Confidence: 49.93% average

Orders: ❌ NOT CREATED

SELECT COUNT(*) FROM orders WHERE account_id LIKE '%paper%';
-- 0 rows

SELECT COUNT(*) FROM ensemble_predictions WHERE order_id IS NOT NULL;
-- 0 (no linkage)

Missing Code: ❌ DOES NOT EXIST

• No file: services/trading_service/src/paper_trading_executor.rs
• No consumer polling ensemble_predictions table
• No order creation logic
• No background task in main.rs

---

Solution

Implement Paper Trading Executor

Architecture:

┌─────────────────────────────────────────────────────┐
│ Background Task (100ms interval)                    │
│                                                     │
│ 1. SELECT FROM ensemble_predictions                 │
│    WHERE order_id IS NULL                          │
│    AND confidence >= 0.60                          │
│    AND action IN ('BUY', 'SELL')                   │
│                                                     │
│ 2. Check risk limits                               │
│                                                     │
│ 3. INSERT INTO orders (...)                        │
│                                                     │
│ 4. UPDATE ensemble_predictions                     │
│    SET order_id = <new_order_id>                   │
│                                                     │
└─────────────────────────────────────────────────────┘

Key Components:

• PaperTradingExecutor struct
• fetch_pending_predictions() - query DB
• execute_prediction() - create order
• create_order() - INSERT into orders table
• link_prediction_to_order() - UPDATE prediction with order_id

---

Implementation Plan

Phase 1: Core (2 hours)

• Create paper_trading_executor.rs (600 lines)
• Implement prediction fetching + order creation
• Add to main.rs as background task

Phase 2: Risk (1 hour)

• Position size calculation
• Risk limits validation
• Circuit breaker integration
• Symbol filtering (reject TEST_SYM)

Phase 3: Testing (1 hour)

• Unit tests
• Integration tests
• End-to-end validation
• Conversion rate monitoring

Total Time: 4 hours

---

Quick Fixes Required

1. Stop Using TEST_SYM

// ml/tests/e2e_ensemble_integration.rs
-let symbol = "TEST_SYM";
+let symbol = "ES.FUT"; // Use real symbol

2. Raise Confidence Threshold

pub const MIN_CONFIDENCE_THRESHOLD: f64 = 0.60; // Not 49.9%

Rationale: 49.93% average is barely above random. 60% filters noise.

---

Success Metrics (After Fix)

Metric	Current	Target
Conversion Rate	0%	>50%
Orders Created	0	>1500
Avg Confidence	49.93%	>65%
Symbols	TEST_SYM	ES.FUT, NQ.FUT, ZN.FUT, 6E.FUT

---

Files to Create

services/trading_service/src/
├── paper_trading_executor.rs   (NEW - 600 lines)
├── paper_trading_config.rs     (NEW - 100 lines)
├── position_tracker.rs         (NEW - 200 lines)
└── main.rs                     (MODIFY - add background task)

---

Validation Commands

After implementation:

# 1. Check orders created
psql postgresql://foxhunt:foxhunt_dev_password@localhost:5432/foxhunt \
  -c "SELECT COUNT(*), symbol FROM orders WHERE account_id LIKE '%paper%' GROUP BY symbol;"

# 2. Check conversion rate
psql postgresql://foxhunt:foxhunt_dev_password@localhost:5432/foxhunt \
  -c "SELECT COUNT(*) as total,
    SUM(CASE WHEN order_id IS NOT NULL THEN 1 ELSE 0 END) as executed,
    ROUND(100.0 * SUM(CASE WHEN order_id IS NOT NULL THEN 1 ELSE 0 END) / COUNT(*), 2) as rate
  FROM ensemble_predictions WHERE ensemble_action IN ('BUY', 'SELL');"

# 3. Monitor paper trading logs
docker-compose logs trading_service | grep "paper_trading"

Expected results:

• ✅ >1500 orders created
• ✅ >50% conversion rate
• ✅ Real symbols (ES.FUT, NQ.FUT, etc.)
• ✅ Predictions linked to orders

---

Detailed Report

See: /home/jgrusewski/Work/foxhunt/PAPER_TRADING_FIX_REPORT.md

Full analysis: Root cause, design, implementation plan, code examples

---

Next Action: Implement PaperTradingExecutor (4 hours)

Priority: 🔴 HIGH - Paper trading pipeline blocked

Impact: Unlocks paper trading execution (3000 predictions waiting to execute)