Fights within 7 days · 🥊 striking edge · 🤼 grappling edge · 🟢 pick is playable at market price · +EV = side beats de-vig market by 5%+ · ⭐ = pick AND value agree
Win probability — each fighter carries a modified Elo rating built from full UFC fight history (K-factor boosted for early finishes, dampened for split decisions). At matchup time the rating is adjusted for reach differential, age, and ring rust (layoff > 9 months), then converted to a win probability with the logistic formula P = 1 / (1 + 10^((Rb−Ra)/400)).
Method of victory — finishing events are modeled as Poisson processes. A KO rate λ is built from the attacker's knockdown output + career KO share, scaled by the defender's chin vulnerability (KO-loss share + striking defense). A submission λ combines sub attempts, the defender's sub defense, and takedown pressure. P(decision) = e^−(λko+λsub) — the chance the fight survives every finishing event.
Value edge — market odds are de-vigged, then compared to the model's probability. A green VALUE tag means the model's number beats the market's implied probability by 5%+ at the best available price. The "play if better than" line converts model probability (minus a 3% safety margin) to American odds — e.g. if it says −120, only bet at −120 or better.
⚠️ Predictions are model estimates, not guarantees. Sample stats ship with this page; live odds load from the Odds Piranha backend when deployed. Bet responsibly — 21+.