The Model Doesn't Care About the Narrative

UFC 329 is built around a rematch the whole sport wants — Conor McGregor back in the Octagon, Max Holloway on the other side of it. The market has made that story very expensive: McGregor sits at 65% implied probability, Holloway at 35%. Our model, which has beaten a coin flip on 2,154 holdout fights but is not validated against betting markets, says the other thing — Holloway 58%, McGregor 42%. That's a 22.9-point gap, the single largest divergence on a trusted bout this card.

That tension — name value and narrative on one side, the underlying numbers on the other — is the thread running through July 11. It shows up in the co-main too, where Robert Whittaker moving up to light heavyweight against Nikita Krylov is priced as the underdog while our model has him winning 60% of the time. It shows up in a bantamweight fight between Adrian Yanez and Cody Garbrandt where the market gap is nearly 28 points. This is a card where the price and the profile keep disagreeing with each other, and that's worth paying attention to — carefully, honestly, and without pretending any of it is a sure thing.

These are the fights where our number and the market's number disagree most, on bouts we actually have data for. What we won't do is call any of them a play. What we will do is tell you exactly why the gap exists.

• Cody Garbrandt (model: 51%, market: 23%, gap: +27.9pp) — The market has essentially written Garbrandt off, and given his recent run, that's understandable. But our model is reading his current sample against Adrian Yanez's current sample and landing almost dead even. The market is pricing a narrative about Garbrandt's decline harder than the numbers in the brief justify. Whether that narrative is right is a genuinely open question — but a 27.9-point gap is a significant one.
• Max Holloway (model: 58%, market: 35%, gap: +22.9pp) — The most visible disagreement on the card. The McGregor name, the return story, the star power — the market is pricing all of it. Our model has zero fight history on McGregor to work with (zero stat-sample fights in the brief), which is its own honest limitation. But what it does see is Holloway's profile: 4.44 significant strikes per minute, a 57% career finish rate, 32 UFC fights of history. The model is suspicious of pricing a fighter this accomplished at 35%.
• Alessandro Costa (model: 56%, market: 33%, gap: +22.4pp) — Cody Durden is the comfortable favorite here. Costa carries a 0.95 knockdowns per 15 minutes rate and a 45% finish rate against Durden's 0 knockdowns per 15 and 33% finish rate. The model sees Costa as the more dangerous finisher and thinks the market hasn't fully priced that.
• Robert Whittaker (model: 60%, market: 47%, gap: +13pp) — Moving up a weight class should cost you something. The market has decided it costs Whittaker the favorite's edge against Krylov. Our model disagrees by 13 points, largely on the strength of what Whittaker brings stylistically (more on that in the matchups section). One fight of sample data is a real caveat here — take the 60% with appropriate humility.
• Benoît Saint Denis (model: 51%, market: 43%, gap: +8.4pp) — The smallest gap on the trusted divergence list, but the profile is interesting. BSD has finished every single UFC fight in his sample — career finish rate of 1.0 — and takes the fight down nearly 4 times per 15 minutes. The market slightly prefers Paddy Pimblett. Our model calls it a coin flip leaning ever so slightly toward BSD.

Ranked by how much we believe the win probability, not by how big the market gap is. The style rates are doing the talking here.

1. Robert Whittaker vs. Nikita Krylov — 60% model confidence. Even with a single UFC fight of data at this weight, what the numbers show is a fighter who brings 3.96 significant strikes per minute with 1.01 takedowns and 1.35 grappling advances per 15 — a complete, mixed game. Whittaker is not walking into this as a raw grappler or a raw striker; he's a problem in both phases. The model believes that blend gives him a real edge at 35% of the time that Krylov's power game ends it early.
2. Max Holloway vs. Conor McGregor — 58% model confidence, with the explicit caveat that McGregor's stat sample is zero fights, so the model is effectively running on Holloway's profile alone on one side. What it sees on Holloway's side is real: 4.44 strikes per minute, 0.9 grappling advances, 0.45 knockdowns per 15, and a 57% finish rate across 32 UFC fights. That's an elite, proven, two-way fighter. The model respects it.
3. Alessandro Costa vs. Cody Durden — 56% model confidence. Costa's 0.95 knockdowns per 15 minutes rate is the standout number in this matchup against a Durden profile that shows zero knockdowns and a 2.23 strikes-per-minute output. Costa finishes nearly half his fights. Durden takes people down 3.41 times per 15 — that's real — but if this stays standing, the model thinks the power edge matters.
4. King Green vs. Terrance McKinney — 53% model confidence. Modest conviction, but the profile holds up: Green at 6.76 significant strikes per minute and 2.39 takedowns per 15 at age 39 is still a high-volume, two-dimensional fighter. The model sees him edging a McKinney who brings the volume (6.25 strikes/min) but whose grappling advances are zero in the sample.
5. Tracy Cortez vs. Wang Cong — 52% model confidence, the closest call on the board. Basically a coin flip with a slight Cortez lean — driven by her takedown output (1.89 per 15) and grappling advances (0.9) against Wang's near-zero grappling rates. Wang is the higher-volume striker (5.93 vs. 4.44), so if this is a pure stand-up fight, it could easily go the other way.

Two categories here: trusted bouts where the model is skeptical of the favorite, and low-data bouts where nobody — including us — should be claiming a real edge.

Market Favorites the Model Questions

• Adrian Yanez (market: 77%, model: 49%) — A 27.9-point gap on a fighter the market has as a heavy favorite. The model sees a near-even fight. Yanez brings 1.05 knockdowns per 15 and an 83% finish rate, so the concern isn't his ability — it's that the market may be pricing him significantly beyond what his recent sample supports against a Garbrandt profile that isn't as dead as the price implies.
• Conor McGregor (market: 65%, model: 42%) — Already covered in the lead, but worth restating plainly: the model has no fight data on McGregor to build from (zero stat-sample fights). It's pricing his opponent's profile and getting a very different number than the market. The market may be entirely right — star power, return narratives, and chin checkpoints all matter in ways models can miss. But the gap is 22.9 points, and you should know it exists.
• Lone'er Kavanagh (market: 66%, model: 48%) — A low-data bout (flagged below), but Kavanagh's market price relative to our number is worth a mention. The model has thin confidence in either direction here.

Low-Data Bouts — The Model Is Blind Here, Defer to the Market

These six fights involve debutants or fighters with one or fewer UFC stat-sample fights. Our model cannot build a meaningful picture. Anyone telling you they have a firm edge in these is working on vibes, not signal — and that includes us.

• Cesar Almeida vs. Damian Pinas (Middleweight)
• Elisha Ellison vs. Gable Steveson (Heavyweight)
• John Garza vs. Farid Basharat (Bantamweight)
• Zach Reese vs. Ryan Gandra (Middleweight)
• Kai Kamaka III vs. Luke Riley (Featherweight)
• Lone'er Kavanagh vs. Brandon Royval (Flyweight)

The Steveson fight in particular — a 92% market price on a fighter with no stat sample — is pure name/narrative. The model literally has nothing to add. Watch the fight; don't build an argument around our number there.

McGregor vs. Holloway — The Southpaw Left Hand vs. The Volume Machine

Conor McGregor comes in southpaw with a 74-inch reach and a career finish rate of 0.8 — historically, when this fight ends early, it ends his way. Max Holloway is orthodox, 69-inch reach, and has never been the power-or-bust type: 4.44 strikes per minute, 0.45 knockdowns per 15, 57% finish rate. What Holloway does is accumulate — round after round, he builds a lead with volume and ring generalship. The five-round format at welterweight theoretically favors that approach. The reach disadvantage (five inches) is real and it matters in the early exchanges. The question is whether McGregor's power closes it before Holloway's volume drowns it out. Zero stat-sample fights means the model can't tell us what the current McGregor looks like — only what Holloway looks like, and Holloway looks like a very, very good fighter.

Whittaker vs. Krylov — The Complete Fighter vs. The Finisher

Nikita Krylov finishes 83% of his fights. That's the headline. He lands 4.29 significant strikes per minute and 0.67 knockdowns per 15 with a four-inch reach advantage (77 to 73) over Robert Whittaker. If Krylov catches Whittaker clean, this ends early — and it has ended early for a lot of people who walked into that left hand. What the model is reading in Whittaker's favor is his completeness: 3.96 strikes per minute with 1.35 grappling advances and 1.01 takedowns per 15. He doesn't need to out-slug Krylov; he can drag the fight into exchanges where his wrestling and his pace become the story. If Whittaker respects the power and mixes his attacks, the model thinks his game is the broader one. One fight of data at this weight class is the honest caveat attached to all of it.

Pimblett vs. Saint Denis — Volume Puncher vs. Volume Grappler

Paddy Pimblett at 5.3 significant strikes per minute and a 71% finish rate is a dangerous man standing. Benoît Saint Denis at 5.6 strikes per minute and 3.93 takedowns per 15 minutes is a dangerous man everywhere — and a 1.0 career finish rate across his four-fight sample means he has finished every UFC fight he's been in. Same reach (73 inches), opposite stances (Pimblett orthodox, BSD southpaw). The orthodox-southpaw angle often produces scrambles and takedown opportunities, and BSD's takedown volume is more than double anything Pimblett's sample shows on defense. The model sees this as roughly even, with BSD's grappling volume being the style edge that nudges it his way.

Yanez vs. Garbrandt — Power Hands Both Ways

Adrian Yanez leads almost every offensive metric in this fight: 4.88 strikes per minute, 1.05 knockdowns per 15, 83% career finish rate. He is a legitimate knockout artist and the market respects that. Cody Garbrandt at 3.13 strikes per minute and 0.3 knockdowns per 15 appears outgunned on volume. But here's the thing: Garbrandt's best version was one of the sharpest, most accurate strikers in the sport, and neither fighter brings meaningful takedown offense (both at 0.67 per 15). This fight almost certainly happens standing. The 27.9-point gap between our number and the market is the model saying it doesn't see the Yanez advantage as quite as lopsided as 77/23. Whether that's the model being smart or naive about where Garbrandt currently is — genuinely hard to know.

Costa vs. Durden — Knockout vs. Takedown

Cody Durden is a wrestler: 3.41 takedowns per 15 minutes, 2.23 significant strikes per minute, zero knockdowns in the sample, 33% finish rate. He wins by controlling position and grinding out decisions. Alessandro Costa is the opposite threat: 3.07 strikes per minute but 0.95 knockdowns per 15 and a 45% finish rate — he's not a volume striker, he's a power striker. Two different paths to victory. If Durden gets this to the mat early and keeps it there, his output advantage on the ground is real. If Costa catches him on the way in — and the knockdown rate says he does catch people — the finish rate suggests he knows what to do with it.

Wang Cong vs. Cortez — Volume Southpaw vs. Takedown Artist

Wang Cong at 5.93 significant strikes per minute is the highest volume striker in this matchup — and in southpaw stance, her angles create problems for orthodox opponents. She lands 0.62 knockdowns per 15, real power for the flyweight division. Tracy Cortez doesn't match that output (4.44 strikes/min) but she changes the fight's geography with 1.89 takedowns per 15 and 0.9 grappling advances. Both fighters are the same reach (65 inches). The style question is whether Cortez can get Wang to the mat consistently enough to neutralize the volume advantage, or whether Wang's pace and power make the sprawl-and-brawl answer irrelevant. The model leans Cortez by a hair — but both profiles have the data to win this differently.

A few profiles worth thinking through for DFS construction — framed by the numbers, not a lineup.

• Brandon Royval — The most eye-catching rate on the card: 8.54 significant strikes per minute in a low-data bout. His DFS ceiling of 126.7 is the highest on the full card. The volume is real and his 57% finish rate gives him score-in-any-round upside. The low-data flag means the model can't validate the matchup — but the output profile is unmistakably a DFS-ceiling asset if you're comfortable with the uncertainty.
• Benoît Saint Denis — 5.6 strikes per minute, 3.93 takedowns per 15, 1.0 career finish rate. The DFS projection of 54.9 leads the Pimblett-BSD fight and his ceiling of 115.6 reflects that finishing ability in every round. The best DFS profiles combine volume and finishes; BSD's sample does both.
• Max Holloway — 54.8 projected points, 114.6 ceiling. The main event format (five rounds) means more scoring chances across more rounds regardless of outcome. Holloway's 4.44 strikes per minute over distance is exactly the kind of floor-building profile that DFS rewards even in decision wins.
• King Green — 54.3 projection driven by 6.76 significant strikes per minute and 2.39 takedowns per 15. That's a high two-dimensional output profile. At 39 years old, the age is a real concern for sustainability, but fight-night output at those rates produces DFS volume.
• Gable Steveson — 72.6 projection, which reflects the market's overwhelming confidence (92%) in a short-format finish. Low-data bout; the model is effectively blind. But if you're building around chalk, the ceiling (119.8) and projection are both significant. Just know you're trusting the narrative, not a validated number.
• Alessandro Costa — 0.95 knockdowns per 15 and a 45% finish rate in a three-round fight against a fighter (Durden) who doesn't carry power. Costa's ceiling of 117.0 is built on the possibility of a finish. His 48.8 projection suggests a floor if it goes the distance. The model's conviction (+22.4pp) is the strongest case for him as a differentiated DFS asset.

Three things to carry into fight night.

The main event price and the model's number are not close. McGregor at 65% implied, Holloway at 35% implied — our model says 42/58 in Holloway's favor. That's a 22.9-point disagreement on a v0, unvalidated model with zero fight data on McGregor to work with. We're not going to tell you what to do with that gap. We are going to tell you it exists, that it's large, and that the model's case for Holloway is built entirely on what he has shown in 32 UFC fights. The market's case for McGregor is built on something the numbers can't fully capture. Both things are true.

Whittaker vs. Krylov is the quietly interesting fight on this card. A 13-point model edge (60% vs. 47% market) in a light heavyweight bout where the model sees a complete two-dimensional game beating a high-finish-rate power striker. If Whittaker's mixed approach limits Krylov's big-swing opportunities, the model thinks he wins more often than the price implies. One fight of sample data is the real caveat — hold it loosely.

Six fights on this card are genuinely low-data situations. Steveson, Garza, Pinas, Gandra, Riley, Royval's side — when the brief has one or zero stat-sample fights, the model is not giving you signal, it's giving you noise dressed up as a number. The market has priced those fights on scouting, reputation, and visual evidence that doesn't live in our data. Defer there. The fights we can actually speak to with some confidence are the ones worth your time and attention — and there are enough of those on July 11 to make this a genuinely compelling card to watch closely.