About MMA Fight Companion

Our model uses real UFC fight data — over 7,000 fights worth of stats. Like any prediction model, it's a probability estimate, not a guarantee. Think of it like a weather forecast: if we say Fighter A wins 65% of the time, that means in 100 simulated matchups they'd win about 65 of them. The model performs best when it has plenty of fight data for both fighters.

All fight statistics are scraped directly from UFCStats.com, the official UFC statistics provider. This includes round-by-round data for every fight: significant strikes, takedowns, control time, and more. Betting odds come from The Odds API, which aggregates lines from major sportsbooks.

Fighter stats update after each UFC event. Betting odds refresh daily. Rankings are updated manually from the official UFC rankings page. Upcoming fight cards sync automatically as they're announced.

Low confidence usually means one of three things: (1) one or both fighters have fewer than 3 UFC fights, so there isn't enough data, (2) the fighters are very closely matched across all dimensions, or (3) there's a clash of styles that makes the outcome highly unpredictable. Low confidence doesn't mean the prediction is wrong — it means the model is honestly telling you it's uncertain.

Edge is the difference between what our model thinks and what the sportsbooks think. If our model gives Fighter A a 60% chance of winning, but the sportsbook odds imply only 50%, that's a +10% edge. Edges above 3% are flagged as potential value — meaning the market may be undervaluing that fighter.

This tool is for entertainment and educational purposes only. It is NOT financial advice. While the Edge Finder highlights discrepancies between our model and the market, sportsbook odds are set by professionals with access to information we don't have (injury reports, training camp intel, etc.). Never bet more than you can afford to lose.

Fighters evolve constantly. A fighter's skills, conditioning, and confidence can change dramatically over 2-3 years. Our model uses recency decay — each fight back is weighted 80% as much as the one before it. So the most recent fight has full weight, the one before that 80%, then 64%, and so on. This captures a fighter's current trajectory rather than who they were 5 years ago.

If a fighter hasn't competed in over a year, they receive an inactivity penalty that scales up to 8% over 3 years. The idea: regular competition keeps timing, reflexes, and cage awareness sharp. Fighters returning from long layoffs historically underperform their career averages.

Not currently. Our model is purely stats-based — it works with what happened inside the octagon. Camp changes, injuries, or tough weight cuts are real factors but require subjective judgment that a data model can't reliably capture. Think of our predictions as a baseline that you can adjust with your own fight knowledge.

The method prediction looks at both fighters' finish tendencies adjusted for the opponent. If Fighter A has a high KO rate but is facing someone with an iron chin, the KO probability drops. We calculate the combined finish rate and distribute it across KO/TKO, submission, and decision based on historical patterns for that matchup profile.