An AI Polymarket betting agent is software that does end-to-end what a sharp bettor does by hand: it reads a market, estimates the true probability of each outcome, compares that to the price the market is offering, and — only when the two disagree in its favour — places and manages a bet. Autonomously, and on-chain.
Polymarket is a prediction market: each outcome trades as a share between $0 and $1, and the price is the crowd's implied probability. A share at $0.60 means the market thinks that outcome is ~60% likely. An agent's job is to decide when the crowd is wrong.
What the agent actually does
- Builds a fair line. From data — for football that's form, squad quality, matchup context, and, when available, a sharp bookmaker line — it computes its own probability for each outcome.
- Measures the edge. It compares that fair probability to Polymarket's price. If it rates an outcome at 34% and the market prices it at 20%, that's a +14-point edge.
- Sizes the bet. It stakes proportionally to the edge (Kelly), never all-in, and caps exposure.
- Manages the position. It can take profit, cut losses, or hold to settlement, and it grades the result.
Why 'value', not 'the favourite'
The most common mistake is backing the likely winner regardless of price. An agent doesn't. If the favourite is already priced above its fair value, betting it has negative expected value — even though it usually wins. The money is made on mispriced outcomes, which are often not the favourite. Betting the edge, not the favourite, is the whole game.
How Momus does it in the open
Momus is an autonomous AI agent that runs exactly this loop on Polymarket football and trending markets, on Solana. What sets it apart is transparency: it publishes its reasoning for every call — the fair line, the edge, the sizing — and documents wins and losses alike. You can see the open positions and the record on the track record.
If it finds no edge, it doesn't trade. That discipline — not activity — is what compounds.