polymarket-arbitrage_skill

Overview

This skill monitors and executes arbitrage opportunities on Polymarket prediction markets. It detects math arbitrage, cross-market mismatches, and orderbook inefficiencies, and includes risk management, P&L tracking, and alerting. Use it to discover actionable edges and to validate automated or manual trading strategies.

How this skill works

The skill scrapes active Polymarket markets, computes implied probabilities, and looks for probability-sum mismatches (math arbs), cross-market price differences, and orderbook execution gaps. It factors taker fees, computes net edge and a risk score, stores state to deduplicate alerts, and can run continuously to notify on new opportunities. Output files capture market snapshots, detected arbs, and alert state for review and auditing.

When to use it

• When scanning Polymarket for guaranteed or near-guaranteed math arbitrage
• When validating viability of an automated trading strategy before automation
• To paper-trade and quantify opportunity frequency and quality
• When monitoring liquidity and orderbook inefficiencies for execution planning
• Before scaling capital allocation from manual testing to automation

Best practices

• Start with paper trading for 1–2 weeks to measure real-world opportunity rate
• Require a minimum net edge (recommend >=3% after fees) and cap position size (max 5% of bankroll)
• Focus on buy-all-outcomes math arbs first; avoid sell-side until experienced
• Log every opportunity and execution and check live orderbook before committing capital
• Use conservative volume filters (e.g., >$100k–$1M) to avoid low-liquidity traps

Example use cases

• Run periodic scans to build a watchlist of candidate markets and their net edges
• Paper-trade detected arbs and record hypothetical P&L to validate the strategy
• Use monitor mode with webhook alerts to notify when new arbs exceed threshold
• Perform micro-tests with $50–$100 to learn execution mechanics and slippage before scaling
• Implement risk rules: daily loss limit, per-opportunity sizing, and automatic alert deduplication