DWCS 91: where our model and the betting market split
DWCS 91 odds vs our model: the books have Kwon Won Il at a coin flip and Martin Kozak at -474. We disagree on three of Tuesday's five Contender Series fights.
Contender Series cards are the one place on the UFC calendar where the betting market and I have almost nothing in common to work with. There is no octagon record to lean on, no ranked opponent to calibrate against, and in most cases no tape that either of us has seen more than once. Everybody is unbeaten, or close to it, and everybody beat someone you have never heard of to get here.
So when the books and our own model look at the same five fights on Tuesday and come out ten points apart on three of them, that is worth a paragraph or two. Dana White's Contender Series 91 runs from the Meta Apex in Las Vegas at 7:00 PM ET, five fights, five contracts on the table, and the disagreements all point the same direction: the market trusts a clean record more than I think it should.
What the two sides actually say
Consensus prices below are de-vigged averages across the sportsbooks we track, taken Monday morning, with the prediction markets left out. Our model number is the AI win probability we publish on each bout page.
| Fight | Market | Our model | Gap |
|---|---|---|---|
| Arlind Berisha vs Quentin Pasley | Berisha 64.1% | Berisha 53.4% | 10.7 |
| Martin Kozák vs Christian Echols | Kozák 77.2% | Kozák 66.6% | 10.6 |
| Kwon Won Il vs Apollo Gomes | Kwon 52.9% | Kwon 63.1% | 10.2 |
Two fights are missing from that table on purpose. On the welterweight and flyweight bouts, one corner has no reach measurement on file, and the model reads a blank field as a zero rather than as an unknown, which throws its size term off a cliff. I am not going to quote a number I know is being fed garbage. Those two get argued the old way, on the record and the odds, in the full card picks.
The one the market is underrating
Kwon Won Il is the most experienced fighter on this card by a distance, and the books have him at a coin flip. He is 14-6 as a pro, he spent six years in ONE Championship, and twelve of his fourteen wins are knockouts. He knocked out Kevin Belingon, who fought four times for the ONE bantamweight title, in the second round in December 2021. In January 2025 he challenged Fabricio Andrade for the ONE bantamweight title at ONE 170 and got stopped in the first round, which is the other half of the same fact: a man who gets a title shot at 29 and a Contender Series slot at 31 has been fighting a level up from everyone else in this building.
He is also here for the second time. On October 14 last year he took his Contender Series audition against Juan Díaz and got knocked out in the second round. Eleven months later he is back at the Apex against Apollo Gomes, a 12-2 Chute Boxe product whose last win was a split decision in Barcelona in March.
The market splits that fight roughly down the middle, Gomes hovering between -115 and +115 depending on the book, Kwon between -145 and -115. Our model has Kwon at 63.1%, and it gets there despite two of its own drivers pointing the other way: it flags Kwon at 31 as past his physical peak and Gomes as inside his, and it likes Gomes' 86% career win rate over Kwon's 74%. The level of the opposition on Kwon's record outweighs both. That is exactly the input a market with no octagon film has the hardest time pricing, and it is why I think the number is wrong.
The two the market is overrating
Martin Kozák is 6-0 with six finishes at 24, all of them in RFA and one Slovak regional show, and the books have installed him at around -474, which is a line you normally see next to a name people have watched fight on television. Our model has him at 66.6% against Christian Echols, the 10th Planet middleweight who is 8-4 but has finished all eight of his wins, seven of them in the first round, including Patrick Downey at Bellator 289. Echols has been stopped three times, which is where the market gets its confidence. He has also never been stopped by anyone whose resume looks like Kozák's, because Kozák has not fought anyone yet whose resume looks like anything.
The main event is the same story in a lower key. Arlind Berisha is 5-0 with five first-round knockouts out of Sarpsborg, Norway, and he is a -210 favorite over Quentin Pasley, who is 3-0 with three first-round knockouts out of Michigan. Berisha's five wins came against opponents who were a combined nothing special in FCR and AFN; Pasley's three came in CES and LFA, which is a step up in room quality if not in opponent quality. Our model splits them at 53.4 to 46.6 and I think that is closer to the truth than a 64% favorite.
Why this keeps happening
The reason is not that the traders are lazy. It is that a Contender Series line has to be made from a record, a highlight package and a weight class, and a record is the loudest of the three. Five knockouts in five fights reads as a knockout artist whether the five men were regional prospects or dentists, and a 14-6 with six ONE losses on it reads as a guy who loses.
Our model does not know anything the traders do not know. It simply weighs the strength of the names on a record more heavily than a human eye does when the human eye has no film to overrule it with. Sometimes that is exactly wrong. On the last Contender Series card, DWCS 90, the model backed two underdogs and the favorites held.
But Kwon at even money is the kind of price that only exists because the shorthand for his career is six losses rather than six years against the best bantamweights outside the UFC. If Dana White is handing out a contract on Tuesday to the man who has already fought at the highest level of anyone in the building, the market will have been the last to know.
Card starts at 7:00 PM ET on Paramount+. Make your own picks on the DWCS 91 predictor, and the results land on the event page as they happen.