The model picked winners at an 83% clip in Perth, but the only fight we actually bet on went the wrong way.
Twelve fights graded, ten winners called correctly. That's 83%, which is a strong Sunday morning for anybody scrolling through results with coffee in hand. The model liked Quillan Salkilld over Mateusz Gamrot in the headliner and got rewarded when Salkilld won by submission. It also nailed Diego Ferreira over Billy Quarantillo, Yadier del Valle over Darren Elkins, Alexia Thainara over Amanda Lemos, Ty Miller over Billy Ray Goff, Steven Asplund over Guilherme Pat, Diyar Nurgozhay over Bruno Lopes, Manoel Sousa over Richie Miranda, Miles Johns over Gianni Vazquez, and Juliana Miller over Ravena Oliveira. Ten correct calls, spread across finishes and decisions alike.
The two misses sting in different ways. Louie Sutherland was the pick against José Montanha, and Montanha ended that one by submission — a clean whiff. The other miss came in the fight we actually had money on, which is the part that actually matters for the wallet.
Only one fight on this card cleared the bar to become a recommended play: Gigi Canuto vs Carol Foro, graded a C. The model backed Canuto. Carol Foro won by unanimous decision. That's a straightforward loss, no asterisks, no bad beat story to tell. Final tally on recommended plays: 0-1, minus 1.00 units.
A single-bet card is a light week no matter which way it breaks, but a loss is a loss. When you've only got one swing and you miss it, the record looks worse than the underlying model performance actually was. The 10-2 winner-picking record and the 0-1 betting record are two different stories, and conflating them does nobody any favors. This card is a clean example of why: elite winner-picking, ugly betting slip.
Here's where the report card gets less flattering. Of the ten fights the model got right on the winner, it also nailed the method of victory in just three of them — Alexia Thainara's decision, Diyar Nurgozhay's knockout, and Manoel Sousa's decision. That leaves seven correct-winner calls where the finish type was wrong. Gamrot-Salkilld, Ferreira-Quarantillo, Elkins-del Valle, Goff-Miller, Asplund-Pat, Johns-Vazquez, and Miller-Oliveira all landed on the right name but the wrong how.
Picking winners at 83% is genuinely good work. Picking how those fights end is clearly a different skill, and on this card it was closer to a coin flip gone bad — 3 of 10, nowhere near the winner-picking rate. If there's a lesson in the finish-method gap, it's that knowing who wins and knowing why they win are not the same muscle, and this card is proof the model's better at the first than the second right now.