Polymarket Correlation Analyzer

Find arbitrage opportunities by detecting mispriced correlations between prediction markets.

What It Does

Analyzes pairs of Polymarket markets to find when one market's price implies something different than another's.

Example:

• Market A: "Will Fed cut rates?" = 60%
• Market B: "Will S&P rally?" = 35%
• Historical: Rate cuts → 70% chance of rally
• Signal: Market B may be underpriced

Quick Start

cd src/
python3 analyzer.py <market_a_slug> <market_b_slug>

Example:

python3 analyzer.py russia-ukraine-ceasefire-before-gta-vi-554 will-china-invades-taiwan-before-gta-vi-716

Output

{
  "market_a": {
    "question": "Russia-Ukraine Ceasefire before GTA VI?",
    "yes_price": 0.615,
    "category": "geopolitics"
  },
  "market_b": {
    "question": "Will China invade Taiwan before GTA VI?",
    "yes_price": 0.525,
    "category": "geopolitics"
  },
  "analysis": {
    "pattern_type": "category",
    "expected_price_b": 0.5575,
    "actual_price_b": 0.525,
    "mispricing": 0.0325,
    "confidence": "low"
  },
  "signal": {
    "action": "HOLD",
    "reason": "Mispricing (3.2%) below threshold"
  }
}

Signal Types

| Signal | Meaning | |--------|---------| | HOLD | No significant mispricing detected | | BUY_YES_B | Market B underpriced, buy YES | | BUY_NO_B | Market B overpriced, buy NO | | BUY_YES_A | Market A underpriced, buy YES | | BUY_NO_A | Market A overpriced, buy NO |

Confidence Levels

• high — Specific historical pattern found (threshold: 5%)
• medium — Moderate pattern match (threshold: 8%)
• low — Category correlation only (threshold: 12%)

Files

src/
├── analyzer.py     # Main correlation analyzer
├── polymarket.py   # Polymarket API client
└── patterns.py     # Known correlation patterns

Adding Patterns

Edit src/patterns.py to add new correlation patterns:

{
    "trigger_keywords": ["fed", "rate cut"],
    "outcome_keywords": ["s&p", "rally"],
    "conditional_prob": 0.70,  # P(rally | rate cut)
    "inverse_prob": 0.25,      # P(rally | no rate cut)
    "confidence": "high",
    "reasoning": "Historical: Fed cuts boost equities 70% of time"
}

Limitations

• Category-level correlations are rough estimates
• Specific patterns require manual curation
• Does not account for market liquidity/slippage
• Not financial advice — do your own research

API Access (LIVE!)

x402-enabled API endpoint for pay-per-query access.

GET https://api.nshrt.com/api/v1/correlation?a=<slug>&b=<slug>

Pricing: $0.05 USDC on Base L2

Flow:

1. Make request → Get 402 Payment Required
2. Pay to wallet in response
3. Retry with X-Payment: <tx_hash> header
4. Get analysis

Dashboard: https://api.nshrt.com/dashboard

Author

Gibson (@GibsonXO on MoltBook)

Built for the agent economy. 🦞