Umar's the Story. Our Numbers Say the Card Is Too.

The headliner is Umar Nurmagomedov against Song Yadong, and the market has made up its mind — Umar is an 83% implied-probability favorite. Our model lands at 63%. That's a real gap, and we'll get into it. But the more interesting tension on this card isn't the main event. It's the undercard, where the model keeps looking at the price and raising an eyebrow.

Sumudaerji against Alex Perez is the biggest trusted divergence on the card — our number is 64%, the market is pricing Sumudaerji at 36%. Denise Gomes is an 83% model number facing a market that has her at 57%. And then there are nine bouts where both the market and us are essentially operating in the dark — debutants, fighters with zero or one UFC fight logged, and sample sizes so thin the model won't pretend it knows anything useful. That's most of this card. Eyes open.

These are the four trusted bouts — enough UFC data for the model to have an opinion — where our number and the market's diverge most. Not picks. Disagreements, with reasons.

• Sumudaerji vs. Perez — model 64%, market 36%, gap of nearly 28 points. Sumudaerji is the model's biggest underdog-flip on the card. His 3.98 significant strikes per minute is meaningful output, he generates 1.45 takedowns per 15 and carries a 7-inch reach advantage on Perez. The market is pricing Perez's finish rate (0.75 career finish rate, and a legitimate 0.97 knockdowns per 15) heavily. The model respects the power but doesn't discount Sumudaerji to 36% on it. Worth watching how the style matchup reads when you sit with it.
• Denise Gomes vs. Xiaonan — model 83%, market 57%, gap of nearly 26 points. An 11-year age gap (26 vs. 37) is part of the model's math. Gomes posts 3.45 strikes per minute against Xiaonan's 3.07, mixes in 2.49 takedowns per 15, and carries a 0.5 career finish rate. Xiaonan has a long UFC résumé and the market respects the name. The model respects the résumé less than the age curve and the stylistic edges Gomes shows in the data.
• Song Yadong vs. Umar — model 37%, market 17%, gap of nearly 20 points. The market has Umar as a massive favorite; our number thinks Song is more live than 17% implies. We're not fighting the result — Umar is excellent — but 37% vs. 17% is the model saying Song's path is real, not theoretical.
• Sean Woodson vs. Jenkins — model 56%, market 44%, gap of nearly 12 points. The smallest divergence in this group. Woodson's 78-inch reach against Jenkins's 68 is a 10-inch advantage, and at featherweight that's a significant framing edge. The model likes him slightly more than the market does. The gap is modest; note it, don't overweight it.

Four trusted bouts, ranked by how much we believe the model's win probability — most believed first. This is the style story behind each number.

1. Denise Gomes (83% model) over Xiaonan. The conviction here is real. Gomes brings a 2.49 takedowns-per-15 rate against an opponent who generates just 1.08 of her own and posts a 0.11 career finish rate — meaning Xiaonan, when she wins, wins decisions. Gomes's 0.46 knockdowns per 15 and 0.5 finish rate say she pushes for results; the model sees an 11-year age gap in a weight class where that matters. Highest model probability on a trusted bout.
2. Umar Nurmagomedov (63% model) over Song. We believe this one directionally — Umar is the right side — but the model just doesn't take him to 83%. His 3.43 takedowns per 15 is elite and is the spine of how he wins fights; Song's 1.49 of his own and 3.65 strikes per minute make him a credible mixed threat, not just a punching bag. The model says Song is live enough to keep this at 63/37, not 83/17. Umar in five rounds is a different proposition than Umar at any shorter distance.
3. Sumudaerji (64% model) over Perez. Almost identical model number to Umar, but the market gap is what makes it notable. Sumudaerji's 3.98 strikes per minute, 72-inch reach (7 inches more than Perez), and 1.45 takedowns per 15 build a well-rounded profile. The concern is Perez's 0.97 knockdowns per 15 — genuine power — and his 0.75 finish rate. If Sumudaerji stays accurate and doesn't let Perez land clean, the model thinks he's the better fighter. The market disagrees by 28 points.
4. Sean Woodson (56% model) over Jenkins. Modest conviction, but the reach edge is real and the model has him as the slight favorite against a market that has it the other way. Woodson's 3.93 strikes per minute with a 78-inch reach is a difficult puzzle to solve. Jenkins switches stances (a legitimate tool) and generates 0.44 knockdowns per 15, so the danger is real. This one is close — 56/44 isn't a declaration, it's a lean.

A couple of things to flag before we get into styles.

On the trusted bouts, the market has Umar Nurmagomedov at 83% and Alex Perez at 64%. Our model is 20 points below the price on Umar and 28 points below on Perez. That doesn't make the market wrong — Umar is a legitimately elite grappler and Perez has genuine knockout power. What it means is the model thinks the price on both has moved past where the data supports it. When the market has priced a fighter to 83% or 64% and your model says 63% or 36%, that's information worth sitting with, even if you ultimately trust the market more. With a v0 model unvalidated against prices, you should.

Then there are the nine low-data bouts. Anyone claiming a real edge here — including us — is guessing. These fights are: Andre Lima vs. Namsrai Batbayar, Kevin Borjas vs. Rei Tsuruya, Kai Asakura vs. Aoriqileng, Francesco Nuzzi vs. Xiao Long, Levi Rodrigues Jr. vs. Liu Ce, Hector Santiago vs. Lawrence Lui, Nilson Rojas vs. Bilal Hasan, Julia Polastri vs. Jingnan Xiong, and Cam Nelson vs. Ding Meng. Between debutants and fighters with zero UFC data, the model is essentially blind. Defer to the market on these and don't mistake the model divergences in that group for information they can't yet be.

Umar Nurmagomedov vs. Song Yadong — Bantamweight, 5 Rounds

Umar averages 3.43 takedowns per 15 minutes. That's the defining number of this fight — elite-level wrestling frequency in a division where most fighters would be happy to hit 1.5. Song posts 3.65 significant strikes per minute and 1.49 takedowns per 15 of his own, so he can mix it up, but his ground offense is largely absent (0 grappling advances per 15). If this goes to the mat, it goes Umar's direction. The wrinkle in five rounds is Umar's 0.38 career finish rate — he wins decisions — and Song's 0.58 finish rate means if Song can keep it standing, he's the more likely finisher. Umar's path is volume wrestling; Song's path is keeping it clean and creating moments on the feet.

Alex Perez vs. Sumudaerji — Flyweight

Alex Perez generates 0.97 knockdowns per 15 minutes — the most dangerous power number among the trusted bouts on this card, and it shows in a 0.75 career finish rate. Sumudaerji answers with 3.98 significant strikes per minute and a 7-inch reach advantage (72 inches to 65), along with 1.45 takedowns per 15 as a secondary threat. This is power vs. volume-plus-range. Perez needs to cut the distance and land; Sumudaerji needs to work behind the reach and not get caught standing flat-footed. Perez's 1.69 takedowns per 15 means he can also go to the mat, but Sumudaerji's output rates hold up either way. The finish rate gap (0.75 vs. 0.17) is why Perez is the market favorite despite what the model thinks about win probability.

Denise Gomes vs. Yan Xiaonan — Women's Strawweight

Gomes (age 26) hits 3.45 significant strikes per minute, generates 2.49 takedowns per 15, and has a 0.46 knockdowns-per-15 rate — she applies pressure everywhere. Xiaonan (age 37) generates 3.07 strikes per minute and 1.08 takedowns per 15, with a 0.11 career finish rate that tells you most of her wins have gone to the cards. The model reads this as a complete fighter against a durable decision-winner with an 11-year age gap. Gomes's double-digit advantage in takedown rate is the clearest edge in the data.

Sean Woodson vs. Jack Jenkins — Featherweight

Woodson has a 78-inch reach. Jenkins has a 68-inch reach. At featherweight, 10 inches is an enormous structural advantage, and Woodson uses it — 3.93 significant strikes per minute with a 0.24 knockdowns-per-15 rate, working from range. Jenkins's switch stance is his equalizer; it makes range-finding harder and he generates 0.44 knockdowns per 15, so the power is real when he gets inside. This fight comes down to whether Jenkins can consistently close the distance against a fighter built to prevent exactly that. Woodson's 0.29 career finish rate suggests he's likely to win by decision if he does win, which matters for DFS ceiling.

Julia Polastri vs. Jingnan Xiong — Women's Strawweight (Low Data)

The model is working with thin data here, so take the style read as directional only. Polastri posts 6.21 significant strikes per minute — the highest volume rate on the entire card, by a wide margin. Xiong, at 38, generates 3.92 per minute with a switch stance. If Polastri's volume translates to UFC pace, this is a significant output edge. The market has Xiong heavily favored (68%) on résumé and experience; the model — even with limited data — likes Polastri's activity rates enough to land at 69%. Classic name-value market price vs. what the numbers actually say.

Kai Asakura vs. Aoriqileng — Bantamweight (Low Data)

Both fighters carry a 0.5 or higher career finish rate, and both post legitimate knockdown numbers — Asakura at 0.82 per 15, Aoriqileng at 0.76. The volume is modest (3.17 and 2.3 strikes per minute respectively) but both are looking to end fights. Low data means the model can't be confident here, but stylistically: two knockout artists with similar build (69-inch reach each), similar ages (32 and 33), fighting a fight that's probably going to get interesting. The model has this coin-flip at 52/48; the market has Aoriqileng at 78%. That's the kind of gap we flag in low-data bouts as a known unknown, not as an edge.

A few profiles worth understanding, regardless of how you build.

Umar Nurmagomedov projects at 111.1 points with a ceiling of 177. That ceiling is the highest on the card among trusted fighters by a significant margin. The reason: 3.43 takedowns per 15 in five rounds generates consistent DFS scoring across rounds even without a finish, and his grappling volume keeps accruing statistics. The 0.38 career finish rate means he's more likely to generate sustained scoring than one explosive moment — which, in a five-round main event, is often more DFS-valuable than a quick finish.

Denise Gomes projects at 77.6 with a ceiling of 130.3. The combination of 2.49 takedowns per 15 and a 0.5 finish rate is what drives the ceiling — she can both accumulate statistics through wrestling and end the fight. At strawweight, where scoring volume is typically lower, her output rates stand out.

Alex Perez projects at 71.4 with a ceiling of 124.3. His is the classic high-variance profile — 0.75 career finish rate and 0.97 knockdowns per 15 means either a big, fast finish or a fight that goes against him. The floor reflects that.

Julia Polastri's 6.21 significant strikes per minute is a volume rate that, if it plays out, generates real DFS accumulation. Low data means uncertainty, but the activity profile is the highest on the card. Volume strikers who actually throw that frequently in the UFC tend to post strong output scores regardless of result.

On the low-ceiling side: Sean Woodson has a 0.29 career finish rate, which caps his upside regardless of how the fight plays. If he wins, it's likely a decision — good for a win bonus, not for a bonus-score ceiling. Factor that in when thinking about his 112.3 ceiling relative to fighters with real finish probability.

The slate here is UFC Fight Night: Nurmagomedov vs. Song · Late, locking at 4:40 AM ET on August 29. Thirteen players available.

One thing to understand about showdown format first: the captain scores and costs 1.5×. So a fighter at $5,200 in a flex spot becomes $7,800 in the captain spot — and every point they score is multiplied by 1.5. That means captain value is about ceiling per dollar at the elevated price, not just raw projection.

Captain spot — Nilson Rojas ($7,800 captain salary): Rojas projects at 84.4 points (flex), which translates to 185.9 captain points at the 1.5× multiplier, against a $7,800 price — 15.9 ceiling points per $1,000. That's the model's top captain option on this slate. Important caveat: Rojas has zero UFC fights logged and Bilal Hasan (his opponent) is also a debutant. The model is essentially blind here — it has no UFC-specific data to work with. The projection is built on thin foundations. The market has Rojas at 85% implied probability; our model says 50% with the same honesty about why. Proceed with full awareness that this is a low-data call.

Flex values:

• Kevin Borjas ($5,000): Projects 66.8 points, 13.4 per $1,000. Low-data bout, so the model is working light, but the salary is at the low end of the pool. The floor is 15 — significant downside risk in a low-data fight.
• Umar Nurmagomedov ($9,000): Projects 111.1 points, 12.3 per $1,000, ceiling of 177. The five-round main event with the highest trusted projection and highest ceiling on the slate. The model has Umar at 63% — a winner — and his 3.43 takedowns per 15 over five rounds is a consistent scoring engine. This is the most trusted projection in the showdown block, on a fighter the model genuinely believes in directionally.

Trap — Bilal Hasan ($9,200): Highest salary on the slate. Model projects 31.3 points — 3.4 per $1,000. The model is below the price here, and the note is straightforward: the model has him at 50% (same as Rojas, for the same reason — no data on either fighter) but prices him as the most expensive option on the board with the weakest scoring projection per dollar. The market has Hasan at 15% implied — making him a significant underdog at the highest price point. That's a combination worth flagging. Whether the market or the model is right about Hasan, the scoring numbers at $9,200 don't add up. This is stated as information from a v0 model, not advice to avoid anyone.

Three things to carry into fight night.

The main event is probably right, but not by that much. Umar Nurmagomedov is a legitimate favorite — his 3.43 takedowns per 15 over five rounds is a difficult problem to solve. But our model has him at 63%, not 83%. Song Yadong is a 0.58 career finisher who can mix in takedowns and strike with volume. The market may be pricing Umar's name and family lineage as much as his current profile. That's not a bet either direction — it's just where the model and the market disagree most clearly on the headliner.

Denise Gomes and Sumudaerji are the undercard names the model most believes. Gomes at 83% with a 26-point market gap, Sumudaerji at 64% with a 28-point gap — both on trusted data with real UFC fight history. The reasons are in the rates: Gomes's grappling volume and age advantage, Sumudaerji's reach and output against Perez's legitimate but volatile power. The model v0 caveat applies everywhere, but these are the bouts where our number and the market diverge most on fights we actually have data for.

Nine bouts on this card are low-data. Honest accounting matters. Between debutants, UFC newcomers, and fighters with minimal sample sizes, the majority of this card is operating in a data desert. The model says so plainly, and anyone pretending otherwise — analytics brands included — is overselling. The four trusted bouts are where the real conversation lives. The rest: trust the market, watch the fights, enjoy the card.