Polymarket V6 - Research-First Architecture

A prediction market scanner with research-first architecture. Fetches markets from Polymarket, researches them via SearXNG, and analyzes with LLM to find trading opportunities.

Architecture

┌─────────────────┐     ┌──────────────────┐     ┌─────────────┐
│ Market Fetcher  │────▶│ Research Engine  │────▶│  Analyzer   │
│ (Polymarket)    │     │ (SearXNG)        │     │ (Ollama)    │
└─────────────────┘     └──────────────────┘     └─────────────┘
        │                        │                      │
        └────────────────────────┴──────────────────────┘
                                 │
                        ┌────────▼────────┐
                        │    SQLite DB    │
                        │  (markets.db)   │
                        └─────────────────┘

Installation

cd /home/max/projects/polymarket-v6
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Usage

Full Scan

Fetch all markets, research new/stale ones, analyze with LLM:

python -m src.scanner --full
python -m src.scanner --full --limit 100  # Test with 100 markets

Incremental Scan

Only process truly NEW markets discovered since last run:

python -m src.scanner --incremental

Maintenance

Update prices, clean resolved markets:

python -m src.scanner --maintenance

View Opportunities

python -m src.scanner --opportunities
python -m src.scanner --opportunities --min-edge 0.10  # Higher edge threshold

Statistics

python -m src.scanner --stats

Configuration

Flag	Default	Description
--db	data/markets.db	SQLite database path
--searxng	http://100.112.76.34:8888	SearXNG URL
--ollama	http://100.112.76.34:11434	Ollama URL
--model	qwen2.5:32b	Ollama model
--limit	None	Limit markets to process
--min-edge	0.05	Minimum edge for BUY signal
--min-confidence	0.6	Minimum confidence

Database Schema

• markets table stores all market data
• Tracks first_seen_at to identify new markets
• Preserves research and analysis on upsert
• Indexes on status, edge, and first_seen

Signal Logic

• BUY: Edge > 5% AND Confidence > 60%
• AVOID: Edge < -10% AND Confidence > 70%
• SKIP: Everything else

Bet Sizing

Conservative sizing based on liquidity:

• Liquidity < $500: max $10
• Liquidity $500-2000: max $25
• Liquidity $2000-10000: max $50
• Liquidity > $10000: max $100

Actual size scaled by edge × confidence.

Key Features

1. Research-First: Web research happens BEFORE LLM analysis
2. No Liquidity Filter: Catches small market opportunities
3. Parallel Research: 200 concurrent SearXNG searches
4. Incremental Mode: Only processes truly new markets
5. Proper Maintenance: Cleans resolved markets, updates prices