By Scott B

Last week, Kneebar News asked ChatGPT to predict the winner of every fight at UFC Mexico City. It turns out the artificial-intelligence chatbot guessed right in an impressive 8-of-12 fights, including flyweight Brandon Moreno’s victory over Steve Erceg in the main event. (See chart.)

But ChatGPT whiffed on its biggest gamble of the night: predicting that middleweight Jose Daniel Medina (+370) would defeat Ateba Gautier (-470), despite the 82% implied probability of a Gautier win. Gautier ended up kneeing Medina in the face to end the fight with 1:28 left in the 1st round.

After the Mexico City broadcast ended, we asked the AI model why it had chosen Medina to upset. It provided three key reasons:

• A belief that UFC veteran Medina had fought tougher opponents than Gautier, a UFC rookie. (According to Tapology, Gautier’s 6 prior victories included an opponent with a 2-23 record)
• The fact that Medina trained at high altitude, giving him a “huge cardio advantage” if he could survive Gautier’s power in the opening minutes
• The mental boost Medina would get from a hometown crowd. (Only problem: Medina is from Bolivia, not Mexico)

“Gautier may have been the favorite for good reason,” ChatGPT explained. “But odds that extreme didn’t reflect the real-world uncertainty of an unproven fighter against a tough opponent in difficult conditions.”

It’s important to note that ChatGPT did not approach the card with a gambler’s eye, such as choosing underdogs it felt were undervalued. Instead, it based its 12 predictions solely on which fighters it thought would win, regardless of the odds.

# # #