Model Picks

We bet four markets: moneyline, decision, over 1½ rounds, and over 2½ rounds. Each one has its own edge requirement and confidence cap, tuned by testing against past fights.

All ROI percentages on this page are backtest results (in-sample, on completed historical fights). They are not a forecast or a guarantee. Forward-paper-traded results — picks logged before the fight happened — will populate after each card. Sustained real-world ROI in sports betting is typically single digits; treat any number above 15% as a backtest artifact until forward data confirms otherwise.

Decision picks need the biggest edge (8%) because the line is sharp. Round-based props work at 5% edge.

Moneyline is the strictest market. A pick is only recommended when four things line up: the model has at least a 3% edgeover the market, it's backing the favorite, it's genuinely confident (its uncertainty band is tight), and our three models — logistic, gradient-boosted, and Bayesian — all agree. Confidence alone isn't enough and edge alone isn't enough; only the combination held up in testing.

We recently added each fighter's pre-UFC record quality — an opponent-adjusted rating of their regional career, so beating a real prospect counts for more than beating a journeyman. That finally made the model reliable on debutants and prelim fighters, so — the confidence check handles fighters the model can't read.