Did the Model Get Song vs. Figueiredo Right?
Yes on the winner, no on the shape of the fight. The model had Song Yadong at 67% confidence, by decision, on the theory that a fresher fighter at his own weight beats a legend who missed weight for his last bout. Song won — but by second-round submission, not the grinding decision the model expected.
Why Did the Model Miss the Method?
The pre-fight read was about pace and volume wearing Figueiredo down over three rounds. Instead, Song found a finish early. That's not a failure of the winner pick, but it's a reminder the model's confidence number is about who wins, not how — and those are separate questions with separate error bars.
How Did the Rest of the Card Go?
6 of 9 decided fights correct (one bout ended in a No Contest and isn't scored either way). A High-confidence pick — Jake Matthews at 84% — landed clean by decision, exactly the kind of result that keeps the model's top tier trustworthy. The misses were all in the Low/Medium tiers, which is where volatility is supposed to live.
Model record this card: 6-3 on decided fights. Main event: correct winner, wrong method.