UFC Fight Night 289 parlay: model picks and the legs we cut

The widest gap between our model and the sportsbooks on tonight's card is not in the main event. It is one fight earlier. Ailin Perez is -161 to beat Norma Dumont, a price that implies a 59.1% win probability. Our model gives Perez 76.1%. That is a 17 point gap, the only one on the card bigger than our own 15 point comfort line, and it is the most interesting leg on either ticket below.

UFC Fight Night 289: Rosas Jr. vs Barcelos goes tonight at the Meta Apex in Las Vegas. Prelims start at 21:00 UTC (2 p.m. Pacific), the main card at 00:00 UTC (5 p.m. Pacific), and the full start times and broadcast guide is here. Twelve fights remain after Mickey Gall vs Sedriques Dumas fell apart; Dumas now meets a late replacement. Every one of the twelve is priced by at least 17 books, and our model has an opinion on all of them.

Before any numbers: this is our model's opinion, published as analysis. It is not betting advice, and nothing here is for anyone under 18.

Two tickets, built from pools that never share a fight. The safe ticket only plays market favorites. The upset ticket only plays market underdogs. The model prices every fight; my job was choosing which of those priced legs to stack, and, more often, which to leave alone. If you want a straight pick on every bout instead, that lives in our full-card predictions.

The safe ticket: six favorites, no coin flips

Photo: Chamatkar Sandhu / CC BY 3.0

Start with the headliner. Raul Rosas Jr. is -156 over Raoni Barcelos, which implies 58.4%. The model has him at 70.6%. The reasons it gives are the tale of the tape: Rosas is 22, Barcelos is 39, and Rosas has absorbed 107 significant strikes in his whole UFC career to Barcelos' 837. Rosas has also won 80% of his first rounds on the judges' cards, Barcelos 27%.

I am not going to pretend the market agrees. This line opened at -180 and has drifted toward Barcelos all week, and there are reasons for that. Barcelos is on a five-fight win streak, ranked one spot below Rosas at bantamweight, and he stuffs 88% of takedown attempts against him. Rosas lands 6.1 takedowns per 15 minutes. This is also the only five-round fight on the card, and the late rounds are where Barcelos has historically done his best work. The model still likes Rosas by a wide margin. It just is not a lock, and I would rather tell you why than hide it.

Then the Perez leg from the top. A 17 point gap is where our rules tell us to suspect ourselves first, so I went looking for a reason the model could be wrong and found the opposite. Perez and Dumont share four opponents: Joselyne Edwards, Karol Rosa, Macy Chiasson and Ashlee Evans-Smith. Perez beat all four. Dumont went 2-2 against the same four, and one of those losses, to Edwards in April, is her most recent fight. The market has been walking toward the model since the open, from -130 to -161. Perez is on a six-fight UFC win streak and averages 4.9 takedowns per 15 minutes to Dumont's 1.4. The warning stays on the leg. So does the leg.

Luis Hernandez over Sedriques Dumas at -245 carries the other flag on this ticket, thin data, and it deserves the disclosure. Hernandez has one fight under the UFC banner, a second-round submission on Dana White's Contender Series on September 15, and he takes this light heavyweight bout on 11 days of turnaround. The model has him at 63.2%, a little below the book's 68.1%. The case for him is mostly on the other side of the cage: Dumas is winless in his last four, a knockout loss, a no contest, then submission losses to Donte Johnson and Jackson McVey. Hernandez is 8-0 and four of his last five wins came by submission.

Photo: Profigrappling / CC BY-SA 4.0

Rodolfo Vieira at -169 over Robert Bryczek is the leg the market has come around on hardest. It opened as a pick'em, -108 each way, and the model had Vieira at 68.0% the whole time. Vieira is an ADCC and IBJJF world champion, and Bryczek's own numbers show the danger: 24% striking defense, 7.67 significant strikes absorbed per minute, and 5.6 minutes of control conceded per 15. The counterweight is not small. Vieira is 37, has lost two straight, including a third-round knockout against Bo Nickal, and Bryczek hits hard enough to have stopped Brad Tavares in the third round last September.

The last two legs are ballast, and I will say so plainly. Montel Jackson at -224 over Ricky Simon is a fight where the model and the market agree almost to the decimal, 66.7% against 66.3%. Both men lost to Barcelos in the last year, Jackson by split decision and Simon by unanimous decision, and Jackson brings a six-inch reach edge and a much lighter damage ledger (217 career significant strikes absorbed to Simon's 829). He has also dropped two straight split decisions, which is why this is ballast and not a value play. John Castaneda at -383 over Heili Alateng is the shortest price on the ticket, and here the model is actually less sure than the books, 71.5% to 76.0%. Both men have been losing, but the one opponent they share split them: Castaneda lost a split decision to Chris Gutierrez, Alateng lost a unanimous one.

Six legs, 10.87x (+987) combined. The math-optimal version of this ticket was five legs at 11.05x, and it had Elves Brener in it. More on him below.

The upset ticket: two dogs where nobody has seen the favorite

Both legs on this ticket come from the card's two fights between UFC newcomers, and both carry the thin-data flag. That is the honest shape of the ticket: it is a bet that the books are overconfident about favorites who have not fought in the Octagon yet, not a claim that the model knows these fighters well.

Melissa Amaya is +177 against Valesca Machado. Amaya debuts at 8-0; Machado's only appearance under the UFC banner is a loss. The model gives Amaya 42.2% against a market number of 34.7%, and it gets there from the pre-UFC records (Amaya unbeaten, Machado with losses on her record before she ever reached the UFC) rather than cage stats it does not have. Two honest problems: the line has drifted away from Amaya all week, from +140 to +177, and the two halves of our model disagree sharply on this fight. It is the least certain opinion on either ticket.

Ilimbek Akylbek Uulu is +234 against Mehemmedeli Osmanli, and this leg is a bet against the price more than against the fighter. The model still makes Osmanli, a 22-year-old debutant at 12-1, the favorite at 58.7%. The books have him at 71.3% with zero UFC minutes on record. Akylbek Uulu (10-3) has real warts, two TKO losses in 2024, but has since won twice by submission and picked up a regional bantamweight title in 2025. The line has not moved all week, -290 at the open and -291 now.

Two legs, 9.25x (+825). That is just under our usual 10x target, on purpose, for the reason in the next section.

What we refused, and why

The math-optimal upset ticket was Amaya plus Brady Hiestand at +296 against Rinya Nakamura, 10.97x. On paper Hiestand is the best value dog on the card, model 40.3% against a market 24.2%. It is also a 16 point gap, and a gap that size is where our rules say the model is more likely missing something than the books are. Here I think I can see what it is missing. Hiestand has not fought since June 2024, 832 days, and his offense is wrestling first against Nakamura, a U23 world champion freestyle wrestler whose UFC takedown defense reads 100%. The line has moved further from Hiestand this week, not toward him. I left him out.

The math-optimal safe ticket used Elves Brener at +107 over Josiah Harrell, which our rules allow because the fight is close to even in the market. I cut him anyway. The model has him at 52.4%, which is a coin flip wearing a number, and he has lost his last three UFC fights. A coin flip does not belong on a ticket called safe, however useful its price would be for reaching the band. Adding him to the upset ticket instead would have built a three-leg, 19x slip out of three thin-data legs, so he is on neither.

Rodolfo Bellato against Christian Edwards is the same story: the model calls it 51.3% to 48.7%, and neither side is a leg. Yazmin Jauregui at -965 over Vanessa Demopoulos adds almost nothing to a ticket's price, and the model is less sure of her than the market (82.0% to 86.8%). The favorites in the two newcomer fights and in Hiestand's fight, Osmanli, Machado and Nakamura, were never candidates for the safe ticket, because the model rates all three well below their prices.

Here are the rules doing the cutting, in plain language. Both corners need at least three UFC fights before we treat a probability as well informed; four of tonight's twelve bouts fail that test. A model-market gap above 15 points is treated as our ignorance, not the book's mistake; two bouts cross it. Every leg needs at least three books pricing it; every fight tonight clears that easily. Probabilities are clipped between 5% and 90% so one confident leg cannot dominate a ticket; nothing on this card came close. Six of twelve fights carry a warning. I played four of those legs, each with the warning named above, and refused two. A model that says "I do not know much about this debutant" out loud is worth more than one that prices him to the decimal and hopes you do not ask.

The honest math

The safe ticket's model probability is 11.0%, against a market-implied 7.13%. That makes the fair price 9.09x against the 10.87x offered, an expectation multiple of 1.20 by the model's own numbers.

The upset ticket's model probability is 17.42% against a market-implied 9.94%. Fair price 5.74x, offered 9.25x, expectation multiple 1.61. Remember what sits under that number: two thin-data legs, one of which the model's own halves disagree on.

Now the caveat that matters more than any figure above. Every combined probability here assumes the legs are independent, and fights on the same card never fully are. They share a cage, a referee pool, a set of judges and a crowd, and one strange night of scoring can move three of them together. Read 11.0% and 17.42% as ceilings on what is plausible, not estimates of what will happen. If you only remember one sentence from this piece, make it that one.

This is our model's opinion, not betting advice. If you bet on any of it, you are 18 or older, somewhere that is legal, and the one deciding how much to risk. The live tickets, with every leg's model and market number, are on the event page, and I will be back after the card to count the damage either way.

Odds as of Saturday, September 26, about 10:40 UTC, several hours before the first fight.