Noche UFC Model Picks: Two Underdogs Worth a Second Look

The single sharpest disagreement between our model and the sportsbooks on tonight's card is not close. Jean Silva is a betting favorite at -414 to beat Jose Delgado in the Noche UFC main event, a price that implies a 77.6% win probability. Our model gives Silva 60.2%. That is a 17.3 point gap, the largest on the entire card, and it is the anchor of the underdog ticket below.

Noche UFC: Silva vs Delgado goes tonight from Desert Diamond Arena in Glendale, Arizona. Thirteen fights, all 24 books we track have priced every one of them, and our model has an opinion on all thirteen. Eight of those opinions made it onto a ticket. The other five did not, and the reasons they got left off are worth more than the picks that made it.

Two tickets, built from disjoint pools. The safe ticket only plays market favorites. The upset ticket only plays market underdogs. Nothing in either ticket is a probability we invented; the model prices the fight, and our job here was picking which of those priced legs to stack.

The safe ticket: six favorites, one honest wobble

Photo: EKapp42 / CC BY 4.0

Anchoring this one is David Martinez at -425 over Dan Ige, the one leg on the whole card where the model is more confident than the market: 82.3% against the book's 77.9%. Martinez is 4-0 in the UFC and carries just 50 career cage minutes to Ige's 274, with a fraction of the career damage absorbed. Ige is 35, on a two-fight skid, and was finished in the first round in February. That is not a coin flip favorite, and the price treats it like one anyway.

The rest of the ticket leans on similar logic without pretending it is all clean. Édgar Cháirez is -198 over Tim Elliott off a three-fight win streak and a submission game that ranks in the 93rd percentile among his weight class, though Elliott has historically dominated first rounds and that is the one wrinkle here. Thomas Gantt (-397) and Regina Tarin (-279) both carry our thin-data flag, Gantt on two UFC fights and Tarin on one, but both are backed by real amateur or professional pedigree (an NCAA Division I All-American wrestler in Gantt's case, an 8-0 pro record with five finishes in Tarin's) and the model still leans their way, just less than the market does.

The leg I'd flag hardest is Brandon Moreno over Joseph Morales at -103, and it needs the disclosure up front: the market has this one basically even (48.8% Moreno, 51.2% Morales), close enough that our own rules let the model's preferred side in even though it is not, technically, the favorite. Moreno is on a three-fight losing skid and that is real. What tips it for me is the common-opponent chain: Moreno is 3-0 with two finishes against Deiveson Figueiredo, the exact fighter who twice knocked out and finished Morales. That is not proof, it is a data point, and I am telling you which one it is rather than dressing it up as more.

Also on the ticket, and the one I trust least: Waldo Cortes Acosta at -190 over Curtis Blaydes. The model has this closer to a genuine coin flip (52.7%) than the -190 price suggests, and Blaydes' own wrestling numbers (4.6 takedowns per 15 minutes to Waldo's 0.4) argue the other way. I kept it for Waldo's recent finishing form and a shared win over Derrick Lewis that Blaydes could not get, but this is the leg most likely to be the one that breaks the ticket.

Combined, six legs price out to 9.50x (+850). That lands just under our usual 10x target, on purpose: the math-optimal sixth leg was Rongzhu as the favorite over Rafa Garcia, and I am not using it, for reasons that will make more sense in the next section.

The upset ticket: two live dogs the market underrates

This is the ticket built around that Silva-Delgado gap. Delgado is +329 to beat Silva, a price implying a 22.4% win chance. The model gives him 39.8%. Delgado is two years younger than Silva, riding a two-fight win streak, and his knockoutArtist profile puts his punch power in the 91st percentile of his weight class. Silva, for his part, is a little over a year removed from being knocked out cold by Diego Lopes, and tonight is the first live test of that chin since. Silva's ranked resume is the real counterweight here (he carries a featherweight top-6 ranking Delgado does not have), which is exactly why the market still likes him. The model just likes him less than the market does.

The second leg is Marwan Rahiki at +136 over Tommy McMillen, model 57.5% against a market number closer to 41%, the second-widest gap on the card. Both corners are 3-0 in the UFC, but Rahiki has finished all three of his UFC wins while McMillen has needed the judges twice, and Rahiki is four years younger. The one number that argues against him: McMillen's own round-1 dominance figures from his past fights dwarf Rahiki's, so this is not a clean sweep of every metric in the data, just the majority of them.

Combined, that is 10.12x (+912) on two legs, both live dogs, both backed by a real, stated reason the model disagrees with the price rather than a hunch.

What we refused, and why

A third value dog, Rafa Garcia at +139 over Rongzhu, actually scored better on the model's own numbers than either leg on the upset ticket, with the model favoring Garcia at 51.8% against a market number in the low 40s. It is not on the ticket because a three-leg version of this parlay prices out around 24x, over the hard 20x ceiling we work under. Quality of legs beats the size of the number, and the discipline here is refusing to pad a ticket past a line we set for ourselves. Garcia's takedown profile (95th percentile in his weight class) against Rongzhu's thin submission defense (6th percentile) is a real style mismatch worth knowing even though it is not a bet you can make through us tonight.

Two fights got cut on price alone. Ignacio Bahamondes over Muslim Salikhov sits at -543, priced so heavily that including it would have added almost nothing to either ticket's multiplier while Bahamondes is on a two-fight skid and the model is noticeably less confident (69.6%) than the market's 81.3%. Yousri Belgaroui over Djorden Santos is worse: -821, the heaviest juice on the card, and here the market (85.9%) is actually more confident than the model (77.65%). When the book is more sure of a favorite than we are, that is not a leg, that is a warning sign wearing a favorite's price.

Sean King's UFC debut against Jessie Rosas got cut on data alone. Zero prior UFC fights on record for King means the model has nothing to hold his number against, and both corners already carry the thin-data flag before you get to that.

The last one is the honest admission: Manon Fiorot over Alexa Grasso, at model 55.3% against a market number of 70.3%, is a real 15-point gap sitting right at our trust boundary. It simply was not needed. The six favorites that made the safe ticket were stronger legs on their own merits, and a max-legs ticket does not have room for a bench player just because the bench player is playable.

The honest math

The safe ticket's model probability is 6.95% against a market-implied 8.45%, for a fair price of 14.39x against the 9.50x offered, which works out to an expectation multiple of 0.66. That number under 1.0 is not a typo and not hidden: six favorites, two of them thin-data flagged and one of them a near coin flip wearing a favorite's price, is not a +EV stack by the model's own math. It is the honest price of playing it safe on a card this uncertain.

The upset ticket is the opposite story: 22.87% model probability against a 9.18% market-implied number, fair price 4.37x against the 10.12x offered, for an expectation multiple of 2.31.

Neither number is a promise. Every combined probability above assumes the legs on a ticket are independent of each other, and fights on the same card never really are. They share a venue, the same set of officials, the same cage, and a crowd that can move a close round. Treat 6.95% and 22.87% as ceilings on what is plausible, not estimates of what will happen. Probabilities are also clipped to a floor of 5% and a ceiling of 90% so no single confident leg can dominate a ticket's math, and every fight tonight was priced by all 24 books we track, so nothing here got excluded for lack of a number to work with.

This is our model's opinion, published as analysis, not betting advice. If you are betting on any of this, you are 18 or older, in a jurisdiction where that is legal, and you are the one deciding how much to risk, not us.

Odds as of Saturday, September 12, roughly 15:22 UTC, a few hours before first fight.