Only fighters who have competed in the last 12 months are available.
FightBrain is a machine learning model trained on 10,000+ UFC fights going back to 2010. Before each event, it analyzes both fighters' stats — striking output, takedown defense, finishing rate, Elo rating, recent form — and outputs a win probability for each matchup.
When the model's confidence meaningfully exceeds what the bookmaker implies (the "edge"), that becomes a pick. The bigger the edge, the higher the stake: LEAN 0.5u · MEDIUM 1u · STRONG 1.5u · LOCK 2u.
Every pick goes live on this site before fight night — publicly visible, never edited after the fact. The record you see is the real record.
Model Performance
Backtested accuracy on held-out UFC bouts. FightBrain Model trained on fighter stat differentials, competition history, and recent form.
What The Model Sees
- Striking output, accuracy & defense differentials
- Takedown offense & defense differentials
- Submission attempt rate differential
- Elo rating (dynamic, updates after each fight)
- Recent form & strength of schedule
- Age, reach, win/loss streak
Top Features by Importance
How much each factor contributed to model predictions. Longer bar = more influential.