polymarket-trading-kit/crypto-trading-bot

This repository is a TypeScript trading bot framework for Polymarket, a prediction market platform where users bet on binary outcomes such as whether Bitcoin will go up or down within a set time window. The bot connects to Polymarket's order book and real-time data feeds, monitors prices, and evaluates when to place trades based on configurable rules. The codebase supports two active strategy modes. The first watches for short-term price momentum in the order book and would buy the UP or DOWN token when the market is trending in that direction. The second watches for situations where both the UP token and the DOWN token are priced cheaply enough that buying both would guarantee a small profit regardless of the outcome. This second approach is what the README refers to as binary market arbitrage. A third strategy is mentioned as private and is not included in the public code. Importantly, the order-placement code for both active strategies is commented out by default. This means the bot runs in observation mode only: it connects to the market, tracks prices, and logs signals, but does not send any real orders unless you manually enable that code. There is also a simulation flag in the configuration file that lets you test the logic without touching live markets. Setup requires Node.js and a Polymarket account with a proxy wallet. You provide your private key and wallet address through environment variables. Strategy parameters such as which cryptocurrency to watch, the time window, entry and exit price thresholds, and position size are all set in a TOML configuration file without touching the code. The project layout separates the WebSocket connection logic, the order-book client, configuration loading, and the trading decision code into distinct folders under a main source directory. The README describes the strategy logic in detail, including the specific conditions that trigger buys and sells for each mode. The description field in the repository metadata is spam text repeated many times, which is worth noting as a sign of low-quality metadata, though the README itself is substantive.