smfworks/smf-kalshi-trader

This repository is an automated trading system for Kalshi, a prediction markets platform where users bet on the outcome of real-world events. Written in Python, it automates the full workflow of scanning markets, detecting trading opportunities, sizing positions, and executing or simulating trades. The system is organized into focused modules: a scanner that fetches market data and records price history, an edge detector that identifies five signal types (two currently live, extreme valuation, where a market price goes above 95% or below 5%, and price inefficiency, where the combined cost to buy YES and NO exceeds $1.05, plus three in progress: momentum, mean reversion, and event correlation), and a sizing module that uses a statistical approach called the Kelly Criterion to decide how much capital to stake. A portfolio module handles both paper trades (simulated, for testing) and live trades. A daily intelligence report is generated by the reporter module. Authentication uses an RSA key-pair against the Kalshi API. The database is SQLite, acting as the single source of truth for market data and trade records. The system is designed to run on a schedule: a cron job syncs price history every four hours and runs the full scan-plus-report daily at 10am. Built-in trading rules enforce tight risk limits: $50 total capital, a maximum of $1 per trade, no more than five open positions at once, and $40 held in a dark reserve. The command-line interface supports a dry-run mode so trades can be previewed before going live.