Medić Finished Rodriguez in 30 Seconds. The Board Had the Wrong Guy.

Thirty seconds. That's how long Uroš Medić needed to end Daniel Rodriguez's night and the main event of UFC Fight Night: Medić vs. Rodriguez. Daniel Rodriguez hit the canvas 30 seconds into Round 1, punches, fight over. Rodriguez came in as the model's 58% favorite. The market had already moved the other way — Medić was the 77% market side — and the market was right. One of the shorter main events you'll see.

The card itself was built almost entirely out of finishes. Of the 14 bouts on the night, 12 ended by KO/TKO or submission, and the other two went to the judges. Every single bout produced a decisive winner — no draws, no no-contests.

Navajo Stirling put Jan Blachowicz away with punches at 2:56 of Round 1 in the light heavyweight co-feature. Blachowicz is a former UFC champion and still a recognizable name at 205; Stirling didn't care. Mateusz Rębecki finished Kyle Prepolec by punches in Round 1 at 4:41 and put up the card's highest DK point total in the process. Bogdan Grad submitted Dennis Buzukja with an arm triangle in Round 2 at 4:33. Robert Valentin caught Duško Todorović in a guillotine choke at 4:14 of Round 1. Gilbert Urbina ended Vlasto Čepo with elbows at 1:01 — barely got his gloves dirty.

Tofiq Musayev knocked out Ludovit Klein in Round 2 at 4:07, which was among the bigger upsets of the night given Klein was the 61% model side and 70% market favorite. Michael Oliveira put away Oban Elliott in Round 1 at 1:49. Nina Milošević knocked out Hailey Cowan at 3:41 of Round 1. Jovan Leka dropped Alexander Poppeck with a kick at 2:22 of Round 1. Stephanie Luciano submitted Marina Spasić via rear naked choke in Round 1 at 3:20. Noah Gugnon finished Milos Janicic with a rear naked choke at 1:21 of Round 1.

The two bouts that went the distance: Borislav Nikolic over Mark Vologdin by decision — judges scored it 29-27 across the board — and Aleksandar Rakic over Marcin Tybura, also by decision, with all three judges at 30-27. Rakic was the market's 75% side; Tybura was the model's 59% side. We'll get to that.

Gate verdict, stated plainly: this is UFC model v0, informational only. The picks gate is closed — there is no historical odds validation, nothing on this card was ever framed as a pick, and nothing here is a won/lost betting record. What follows grades the accuracy of the published probabilities. That's it.

The board went 8-of-14 on the night — 57.1% accuracy, logloss 0.6936, Brier 0.2502. The market, where published prices were available, went 12-of-14: 85.7% accuracy, logloss 0.4884, Brier 0.1551. The market was meaningfully sharper on every metric. That's the honest number, and it's on the record.

Three divergence flags of 8 percentage points or more were journaled before the event — spots where the model and market disagreed materially. The model's side won zero of the three.

• Daniel Rodriguez vs. Uroš Medić: 34.9 pp divergence. The model's side was Rodriguez. Rodriguez lost in 30 seconds.
• Navajo Stirling vs. Jan Blachowicz: 21.5 pp divergence. The model's side was Blachowicz. Blachowicz was finished in Round 1.
• Marcin Tybura vs. Aleksandar Rakic: 33.7 pp divergence. The model's side was Tybura. Tybura lost a unanimous decision.

Three flags. Three losses. The market had all three right. At 57.1% accuracy against an 85.7% market benchmark, the v0 model was below the bar this card needed it to clear. That's the receipt.

The realized finish split: 8 KO/TKOs, 4 submissions, 2 decisions across 14 bouts. Twelve of 14 ended before the scorecards — that's the kind of card that plays well on the highlight reel and badly for anyone rostered to a fighter who got finished early.

The eight KO/TKOs came via punches in six bouts, elbows in one (Urbina over Čepo), and a kick in one (Leka over Poppeck). The four submissions were a guillotine choke, an arm triangle, and two rear naked chokes. A note on method grading: per-bout finish-method probabilities were never published on the board — the sim journals DK-point bands only — so realized methods are reported here as results. There is no per-bout method number to grade them against.

On the sim layer: the board journaled p10/median/p90 bands for 28 fighter-slots. Actual DK scores landed inside those bands for 67.9% of fighters — below the sim's validated 75–85% walk-forward coverage target. The mean projection across the card was 59.0 points; the mean actual was 58.3. A projection bias of 0.8 points — nearly no systematic lean either direction on average. The problem wasn't the center; it was the tails. Nine fighters busted their bands, and you'll find their names in the next section.

The top of the DK leaderboard on this card belonged to fighters who finished early and convincingly.

• Mateusz Rębecki — 143.7 DK points, pre-event projection 88.2, band 26.9–125.2. Above ceiling by 18.5 points. The card's high score, and a band bust.
• Uroš Medić — 129.5 DK points, pre-event projection 65.0, band 6.4–116.3. Above ceiling by 13.2 points. Finishing the main event headliner in 30 seconds will do that.
• Bogdan Grad — 120.1 DK points, pre-event projection 73.8, band 9.8–121.8. Inside the band. One of the cleaner outcomes on the sheet.
• Michael Oliveira — 116.8 DK points, pre-event projection 77.2, band 17.5–118.2. Inside the band.
• Gilbert Urbina — 113.9 DK points, pre-event projection 40.5, band 3.5–102.3. Above ceiling by 11.6 points. The board had him projected at 40.5 — he more than doubled it. The market had him at 26% — a significant underdog who didn't act like one.
• Nina Milošević — 108.4 DK points, pre-event projection 80.5, band 19.0–118.6. Inside the band.

Three of the six top scorers covered their ceiling. The other three — Rębecki, Medić, and Urbina — each finished faster and more completely than the sim anticipated.

No spin here. The board's biggest_misses field returned empty — which means the named misses live in the divergence flags and the band busts, and there are plenty of both.

Divergence Flags — 0 for 3

The three spots where the model broke hardest from the market, the model was wrong on all three. Rodriguez (58% model side, lost in 30 seconds). Blachowicz (model side, finished Round 1). Tybura (59% model side, lost a clean 30-27 decision). When the model and market diverged by more than 8 percentage points on this card, take the market. That's not a prescription going forward — it's just what happened on August 1.

Band Busts — 9 fighters outside the p10–p90 window

Biggest ceiling bust: Mateusz Rębecki — actual 143.7 DK points against a ceiling of 125.2, over by 18.5. Early finish, high-volume, the scoring exploded past the sim's p90.

Biggest floor bust: Vlasto Čepo — actual 2.8 DK points against a floor of 11.4, under by 8.6. Finished by elbows in 61 seconds — the sim's p10 assumed he'd be more active than that.

The full list of band busts, with the precomputed miss margins:

• Uroš Medić — above ceiling, miss by 13.2
• Gilbert Urbina — above ceiling, miss by 11.6
• Tofiq Musayev — above ceiling, miss by 1.2
• Daniel Rodriguez — below floor, miss by 3.3 (finished in 30 seconds, scored 0.4 DK points against a floor of 3.7)
• Jan Blachowicz — below floor, miss by 6.0
• Milos Janicic — below floor, miss by 3.0
• Alexander Poppeck — below floor, miss by 0.8

That's 9 of 28 banded fighters outside the window — band coverage of 67.9% against a 75–85% target. Early finishes drove most of the floor busts; two fighters (Rębecki, Medić) scored at a level that outran the ceiling. That's a tail problem more than a center problem, and it's on the record.

Three things this card taught, and they go on the board permanently.

Medić is the real deal at welterweight. The market knew it — 77% — and Medić proved it in 30 seconds. Whether the model catches up to what the market already priced in is v0's central challenge. When the divergence is 34.9 percentage points and the model's side goes down in half a minute, the model needs to ask harder questions about what it's weighting.

When the model and market disagreed materially on this card, the market was right every time. Three flags, three model-side losses. That's not a pattern you ignore. The v0 Brier of 0.2502 against a market Brier of 0.1551 is a real gap — not a rounding error. The work to close it starts with understanding why the divergence flags all broke the same direction.

On the sim: 67.9% band coverage on a card with 12 finishes is the expected outcome of a violent night, not a modeling failure in isolation. The projection bias was 0.8 points — essentially zero systematic lean. But the tails were wide and wild: fighters either obliterated their ceiling (Rębecki, Medić, Urbina) or cratered through the floor (Rodriguez, Blachowicz, Čepo). Early finishes compress loser scores to near-zero in ways that even a well-calibrated p10 can't fully absorb. That's the honest explanation — and it doesn't make the busts disappear from the record.