Salkilld Chokes Out Gamrot, and a Wild Card Goes 12-for-12

The event was named for Mateusz Gamrot. The night belonged to Quillan Salkilld. Four minutes and twenty-five seconds into the first round of their lightweight main event, Salkilld locked in a rear naked choke and Gamrot was done. Both the board (53%) and the market (56%) had Salkilld as the favorite — slim, but real — and he delivered with the kind of finish that turns a slim favorite into a statement.

The card around it was genuinely fun. All 12 bouts produced a winner — seven by finish (four KO/TKO, three submissions), five by decision. Yadier del Valle needed exactly 35 seconds to stop Darren Elkins in featherweight. Diyar Nurgozhay knocked out Bruno Lopes at 4:59 of the first round — one second from the bell, one very long night for Lopes. Ty Miller found Billy Ray Goff at 0:15 of Round 3, fifteen seconds into the final frame, to end a welterweight fight that went the distance in time but not in drama.

On the submission side beyond the main event: José Montanha cranked out Louie Sutherland via neck crank at 1:50 of Round 1, and Juliana Miller rear-naked-choked Ravena Oliveira in the second. The decisions — Steven Asplund over Guilherme Pat, Diego Ferreira over Billy Quarantillo, Manoel Sousa over Richie Miranda, Alexia Thainara over Amanda Lemos, and Carol Foro over Gigi Canuto — filled out a full night of fights where nobody left without a result.

Miles Johns punched out Gianni Vazquez in Round 1 at 3:09 to round out the finishes. Start to finish, this card moved.

Ground rules first, every time: the UFC model is v0 — informational only, picks gate closed. This is a model-accuracy receipt. Nothing on this card was ever framed as a pick, and nothing here is a won/lost betting record. We grade the journaled probabilities honestly, full stop.

On 12 bouts, the model went 75% accurate (logloss 0.6569, Brier 0.232). The market — where published prices existed — went 83.3% accurate (logloss 0.5377, Brier 0.1757). The market was sharper on this card. That's the honest number and it goes on the record.

No divergence flags of 8 percentage points or more were logged between the model and market on any bout, and the biggest-misses list came back empty. That means the board wasn't dramatically wrong in a single place — it was just a tick behind the market across the board. No one whiff to point at; just a calibration gap the v0 label already acknowledges.

One important asterisk: 10 of 12 bouts were flagged low-data — debutants or near-debutants where the model was blind by design and deferred to the market. Grade those exactly as that. The two full-data bouts were the Gamrot-Salkilld main event and the Quarantillo-Ferreira co-feature. The board had both winners favored, and both won.

How the card actually ended: 4 KO/TKO, 3 submissions, 5 decisions. All 12 bouts were decisive — no draws, no no-contests.

The sim journaled DK-point bands (floor = p10, ceiling = p90) for all 24 fighters. Of those 24 bands, 17 of 24 contained the actual result — band coverage of 70.8%. The sim's own walk-forward target is 75–85%, so this card came in a few ticks light. Nothing alarming, but noted.

On projection bias: the sim's mean projection was 61 DK points per fighter; the mean actual came in at 63.6, a bias of -2.5 points. The model was very slightly conservative — it slightly underestimated how much scoring was coming. That's in the noise, but the direction is worth tracking as v0 matures.

One thing the board never published: per-bout finish-method probabilities. The sim journals DK-point bands only. So the KO/TKO and submission results above are reported as exactly what they are — results. There is no per-bout method number to grade them against, and we won't pretend otherwise.

The top of the DK scoring sheet was owned by three fighters who all beat their published ceilings — and one who landed neatly inside her band.

• Diyar Nurgozhay — 129.4 DK points. Pre-event projection: 64. Published band: 12.8–110.2. He went 19.2 points above the ceiling. A late first-round KO will do that.
• Yadier del Valle — 127.3 DK points. Pre-event projection: 81.4. Published band: 25.2–113. He cleared the ceiling by 14.3 points. Thirty-five seconds, one punch, maximum leverage.
• Ty Miller — 126 DK points. Pre-event projection: 35.8. Published band: 2.3–101.2. He went 24.8 points above ceiling — the biggest ceiling bust on the card. The board projected him in the low-to-mid range and he went out and finished the fight in the third. Nobody saw that coming at that scale.
• Alexia Thainara — 109.9 DK points. Pre-event projection: 92.1. Published band: 19.1–143. Covered. A decision win that scored well — she landed inside the band and was one of the cleaner outcomes for the sim.
• Miles Johns — 104.9 DK points. Pre-event projection: 66.2. Published band: 8.7–115.3. Covered. The first-round KO landed within the band, though well above the median projection.
• Quillan Salkilld — 104.7 DK points. Pre-event projection: 98.2. Published band: 11.7–181.4. Covered. The main event winner tracked almost exactly to projection — a clean result in a wide band.

The three fighters who topped the leaderboard all blew past their ceilings. If you rostered them, you got there with upside the board didn't fully account for. If you faded them because the ceiling looked low — well, that's the ceiling problem the misses section covers.

No biggest-misses on the model side and no divergence flags — but the sim band busts are where this card bit back, and we name them plainly.

Seven of 24 fighters busted their bands — three above the ceiling, four below the floor.

Above Ceiling

• Ty Miller — actual 126 DK points, ceiling 101.2, over by 24.8. Biggest ceiling bust on the card. The sim had him with a median projection of 35.8 and a ceiling just over 100. He nearly doubled the ceiling.
• Diyar Nurgozhay — actual 129.4, ceiling 110.2, over by 19.2. A first-round knockout at 4:59 generates a lot of points in a hurry.
• Yadier del Valle — actual 127.3, ceiling 113, over by 14.3. Thirty-five seconds and it was over — the speed of the finish pushed him past the top of the band.

Below Floor

• Louie Sutherland — actual 1.5 DK points, floor 13.8, under by 12.3. Biggest floor bust on the card. He got neck-cranked in under two minutes and scored almost nothing. The sim had a floor that assumed more fight time than he got.
• Mateusz Gamrot — actual 1.8 DK points, floor 3.7, under by 1.9. The event's co-headliner got submitted in Round 1 and scored next to nothing — even a wide band (3.7–161.7) couldn't hold a 1:8 outcome at the very bottom.
• Gianni Vazquez — actual 2.8 DK points, floor 6.4, under by 3.6. First-round TKO loss, minimal scoring.
• Darren Elkins — actual 0 DK points, floor 1.6, under by 1.6. Del Valle ended it in 35 seconds and Elkins scored a zero. Hard to model for, impossible to soften.

The pattern on the floor busts is consistent: fighters who lost quickly in Round 1 scored below where even the conservative p10 floor had them. The pattern on the ceiling busts is equally consistent: fighters who finished quickly in Round 1 blew past the top end. Fast finishes in either direction are where the sim's bands get stretched the most. That's on the record.

Three things this card put on the record:

1. Quillan Salkilld is the real thing. He was a slim favorite in a fight that bore his opponent's name, and he submitted a legitimate lightweight inside five minutes. That's a meaningful statement — not a fluke, not a gift. The board had him at 53%, the market at 56%. He covered both with authority.
2. Fast finishes are the sim's hardest problem. Every band bust in both directions on this card traces back to the same root cause: a fight that ended very quickly. When a fighter gets finished in under two minutes, they score near zero and blow through the floor. When a fighter delivers a first-round KO, they score enormous and blow through the ceiling. The bands are built on a wide range of scenarios, but a 35-second knockout or a 1:50 neck crank compresses that distribution to its extreme. The sim knows this — the 70.8% coverage on a card full of quick finishes reflects exactly that tension.
3. The model is v0 and the market beat it clearly. 75% to the market's 83.3%, with worse logloss and a worse Brier score. Ten of twelve bouts were low-data, which constrained what the model could do — but the market had the same information problem and navigated it better. The gap is real, it goes on the record, and closing it is the work.