A model scores every UFC fight from history alone, without seeing the market. Where its number disagrees with the price, that gap is the edge this desk trades. Every bout is on the board, not only the picks.
Polymarket quotes both fighters. The price is the crowd's estimate of the chance each one wins.
A model built from past bouts estimates the same chance without seeing the price. Fighters with too little history score as no coverage rather than a guess.
Where the model is more confident than the market, that gap is the edge.
Six checks can block a pick — thin fighter history, a price too near even money, an excluded weight class, and three more. A blocked bout stays on the board with its guard shown.
The public market browser remains available while private analytics reconnects.
An edge is not a prediction. The claim is that the price is wrong, not that the fighter wins. Most picks lose.
The history is thin. Fighters compete a few times a year, so even a long career is a small sample — which is why no coverage is usually the biggest bucket on the board.
The guards were chosen, not fitted. A guard set too strict quietly costs picks and leaves no trace beyond the guarded count.
Prices move after a scan. An edge shown here may already be gone; graded results live in History.