Three arbitrage bot types
A trader outlined three distinct automated arbitrage bot designs — a convergence bot, a mispricing detector, and a compounder that rebalances yields — giving a quick taxonomy you could prototype against multiple markets. (x.com)
Arbitrage bots automate a simple trade: buy the cheaper side of the same payoff and sell the richer one before the gap closes. On prediction markets and crypto venues, traders now describe three common versions — convergence, mispricing, and yield-rebalancing bots. (docs.polymarket.com) A convergence bot watches prices that should add up cleanly and trades when they do not. On Polymarket, each outcome is priced between $0 and $1 on a central limit order book, so a YES share at 48 cents and a NO share at 50 cents imply a 98-cent bundle that still pays $1 at settlement. (docs.polymarket.com) A mispricing detector scans related markets for contradictions instead of one market for a broken sum. One open-source builder described buying YES on Kalshi at 35 cents and NO on Polymarket at 63 cents, a 98-cent package tied to the same event that pays $1 when the result is final. (dev.to) A compounder starts with the same small spreads but focuses on turning capital over faster. The same builder said holding a 2% trade until a market expires can tie up money for months, so his bot rotates out and redeploys capital instead of waiting for maturity. (dev.to) These three designs map to three different problems in automated trading. One fixes prices inside a single market, one compares prices across markets or venues, and one reallocates cash between positions to keep earning the highest available return. (github.com) The setup matters because prediction markets publish machine-readable order books and application programming interfaces that bots can query continuously. Polymarket says its order book is public and its official TypeScript, Python, and Rust clients support market data, authentication, and order management. (docs.polymarket.com) The economics are tighter than the examples make them look. A recent Polymarket-focused trading guide said spreads generally need to clear roughly 2.5% to 3% after fees and gas costs to remain worth taking, and the platform’s fee rules vary by market type. (docs.polymarket.com, quantvps.com) Cross-market bots also need to normalize different plumbing before they can compare prices. Polymarket uses a hybrid decentralized order book on Polygon, while Kalshi runs a federally regulated event exchange where contracts trade from 1 to 99 cents and settle to $1. (docs.polymarket.com, kalshi.com) That is why traders keep reducing the field to a short taxonomy. If you can name whether a bot is chasing convergence, mispricing, or compounding, you can usually tell what data it needs, how fast it must trade, and where the real risk sits. (github.com, dev.to)
