Every Momus bet starts from a single question: what is this outcome actually worth, and is the market paying more than that? The answer is a fair probability, and the gap between it and Polymarket's price is the edge.
1. Build a fair line
For a football match, Momus builds its fair probabilities from the data — recent form, goals for and against, scoring and clean-sheet rates, matchup context — and, when a sharp bookmaker line is available, de-vigs it and blends the two. When there's no sharp market, it leans on the model and its own read, and treats that call as higher-variance.
2. Measure the edge against the market
Polymarket's price is an implied probability. Momus compares its fair line to that price for each outcome. A real edge looks like: fair 34%, market 20% — the crowd is underpricing it by 14 points. Crucially, it measures the edge against the blended fair value, not against whoever is the nominal favourite.
3. Only bet the mispricing
This is where discipline shows. A team can be Momus's slight favourite in the read and still be a pass, because the market has already priced it above fair. The value is wherever the market is most wrong — often the underdog. If nothing is mispriced past a threshold, Momus takes no bet at all.
4. Size with Kelly, cap the risk
Stake scales with the size of the edge (a fractional Kelly), never the whole bankroll, with per-bet and total-exposure caps. Bigger edge, bigger stake; thin edge, small stake or none.
5. Grade and calibrate
Winning isn't enough — Momus checks calibration: does a 70% call actually win about 70% of the time? A calibration layer continuously tunes its stated probabilities to the real results, so the edge stays honest as the sample grows. It's the discipline most tipsters skip.
Every step is published. See it applied match-by-match on the track record, or read what an AI Polymarket betting agent is.