Polymarket Kalshi Limitless Triangular Arbitrage

Most published cross-platform prediction arb bots cover two venues (Polymarket vs Kalshi). connectcomputers/arb-bot adds Limitless as a third venue, which meaningfully expands the opportunity set because Limitless often has stale quotes on niche markets that already tightened on the bigger venues. The strategy is straightforward: for each event with matching resolution criteria across all three venues, compute the implied probabilities and look for a combination where buying YES on one venue and NO on another (or a complete set across venues) yields a guaranteed spread after fees.

The bot first builds a market-matching layer (title normalization, resolution-source matching, date alignment) between Polymarket's Gamma API, Kalshi's REST API, and Limitless. Once matched, it computes the cross-venue order book and flags violations of the no-arbitrage bound (sum of YES prices < 1 or > 1). Execution uses FOK or IOC orders on each leg with a pre-computed hedge ratio; the risk is leg-in risk if one side fills and the other does not.

Key implementation concerns: (1) resolution criteria mismatches are the biggest silent killer — always diff the rules text, not just the title; (2) each venue has different fee structures (Kalshi's taker fee is formula-based, Polymarket is currently zero on most markets but this can change, Limitless has its own schedule); (3) withdrawal/settlement timing differs, so capital efficiency must be modeled across the whole cycle, not just per-trade.

The repo is small (~762 KB, 0 stars) with a GARBLED_USERNAME flag, so it is a hobbyist implementation. Do not trust it with real capital without independently verifying the matching logic and fee accounting. It is best treated as a starting scaffold rather than a turnkey system.

Risks: venue API changes, resolution disputes, and capital lockup on the losing venue's settlement schedule. The tri-venue structure also means more leg-in risk than a two-venue arb.