UFC 334: Gane vs Hokit: Model Analysis

Event: UFC 334: Gane vs Hokit Posted: November 11, 2026 · VAR Research First fight: November 15 2026, 02:00 UTC Card scope: 13 fights, lines locked at posted time Methodology: Methodology Model accuracy: Performance

Where the model diverges from consensus

Two matchups show the largest gap between our projected win probability and the public consensus.

Charles Radtke vs Stephen Thompson. Public consensus implies Charles Radtke at 43.2%. Our model projects 66.0%. A 23-point gap, with Charles Radtke materially underrated by the market.

Chris Duncan vs Drew Dober. Public consensus implies Chris Duncan at 61.5%. Our model projects 79.3%. A 18-point gap, with Chris Duncan materially underrated by the market.

Full card projection

No projection is published for 1 bout: our database is missing career statistics for a fighter (Bilal Hasan), and a number built on blank inputs would not be a real projection. These bouts are left out of the card record below.

Matchup notes

Ciryl Gane vs Josh Hokit. Ciryl Gane’s striking volume holds the advantage in a clear lane mismatch.

Kayla Harrison vs Amanda Nunes. In a wrestling-heavy matchup, Kayla Harrison’s superior takedown defense (100% vs 82%) projects to dictate position. Kayla Harrison absorbs notably less (1.1 vs 2.6 per minute), projecting better damage profile through three rounds.

Caio Borralho vs Yousri Belgaroui. Caio Borralho enters 5-0 in his last 5; Yousri Belgaroui is 3-0. Recent-form trajectory tilts the matchup.

Uros Medic vs Kevin Holland. Striker-vs-striker matchup with Uros Medic carrying the volume edge (5.6 vs 4.3 sig strikes per minute). Uros Medic outpaces Kevin Holland on strike volume (5.6 vs 4.3 per minute).

Stephen Thompson vs Charles Radtke. Charles Radtke’s takedown defense (100%) is the largest gap on the board against Stephen Thompson’s (63%).

Drew Dober vs Chris Duncan. Chris Duncan’s grappling pressure (3.3 takedowns / 15 min) projects to neutralize Drew Dober’s striking lane. Chris Duncan holds the takedown-rate edge (3.3 vs 0.6 per 15 min).

Jim Miller vs Terrance McKinney. Terrance McKinney outpaces Jim Miller on strike volume (6.2 vs 2.9 per minute).

Donte Johnson vs Cesar Almeida. Donte Johnson holds the takedown-rate edge (6.0 vs 0.0 per 15 min).

Jared Gordon vs Melquizael Costa. Melquizael Costa absorbs notably less (3.0 vs 4.1 per minute), projecting better damage profile through three rounds.

Macy Chiasson vs Bia Mesquita. Bia Mesquita holds the takedown-rate edge (4.2 vs 1.9 per 15 min).

Juan Diaz vs Alatengheili. Juan Diaz outpaces Alatengheili on strike volume (5.8 vs 2.8 per minute).

Evan Elder vs Samuel Sanches. Style symmetry between Samuel Sanches and Evan Elder. The model declines to favor either side beyond noise.

How to read this

Projections come from VAR’s production UFC ensemble model, refreshed nightly. The model combines fighter performance metrics, simulation-based fight modeling, and historical opponent-strength signal, with conservative treatment of fighters who have limited recent data.

Confidence labels reflect projection spread:

• High: pick probability above 65%
• Medium: pick probability between 55 and 65%
• Coin flip: pick probability below 55%

Note: this analysis does not incorporate fight-week intelligence such as late injury reports or weight cut difficulty. Those signals are tracked through a separate process.

Model accuracy — live forward record

Every fight the production model priced before the bell since 2026-05-30, graded against the official result (n=285, through 2026-11-08; as of 2026-11-11):

• Straight-up winner accuracy: 62.8% (95% interval [57.1%, 68.3%])
• On the 275 fights with a pre-bout market price: the market’s favorite won 70.5%; Brier score model 0.232 vs de-vigged market 0.201 — the market is better calibrated.
• No betting tier has a validated positive return. VAR does not present UFC as a betting edge.

Full record, per card, with the withdrawn validation noted: Performance.

Results

Populates after card concludes.

For research and informational purposes only. Probabilities reflect model output and may diverge from realized outcomes.