Rodrigues Took the Belt Test and Passed. The Board? Mostly.

The card was named for Anthony Hernandez, and the card humbled him. Gregory Rodrigues went the full five rounds in the middleweight main event and walked out with the decision — scorecards reading 47-48, 47-48, 46-49 — despite both the model (58%) and the market (61%) installing Hernandez as the favorite. Rodrigues finished with 135.6 DK points; Hernandez finished with 65.6. That's your headline.

The rest of the night was a finish festival. Five of the seven completed bouts ended by KO/TKO, and the card wasted very little time getting there.

Reinier de Ridder gave the co-main event crowd its cleanest moment, stopping Roman Dolidze by punches at 4:01 of the first round in their light heavyweight bout. Quick, authoritative, and on the right side of the market favorite line at 73%.

Shamil Gaziev continued his recent habit of making heavyweight look uncomplicated, knocking out Kennedy Nzechukwu by punch at 1:20 of the first round. The model had Nzechukwu as a 51% favorite; the market had Gaziev at 55%. A coin flip decided by a heavy hand.

Vitor Petrino was the board's most confident call on the finished bouts — 73% model, 60% market — and he delivered a clean 27-30 on all three judges' cards against Serghei Spivac at heavyweight.

The women's strawweight card gave us two finishes. Shanelle Dyer, the board's heaviest favorite of the night at 91%, stopped Elise Reed by punches at 1:42 of round three. Carli Judice was even faster — she finished Jeisla Chaves by kick at 1:39 of the first round.

Jackson McVey rounded things out with a first-round finish of Wes Schultz via knee at 4:13 in a middleweight bout flagged as low-data — the model deferred to market pricing there by design.

One bout — Gauge Young vs. Kody Steele at lightweight — has no result in the record. Cancelled or unreported; it doesn't grade.

First, the framing that has to come first: this is a v0 model operating in informational mode with the picks gate closed. No historical odds validation has been completed. What follows grades the accuracy of the journaled probabilities — a receipt, not a record.

On the seven gradeable bouts: the model went 5-of-7 (71.4% accuracy, logloss 0.5202, Brier 0.1737). The market, where published, went 6-of-7 (85.7% accuracy, logloss 0.4702, Brier 0.1476). The market was sharper on this card. That's the honest read.

Three of the seven bouts carried the low-data flag — McVey/Schultz, Reed/Dyer, Chaves/Judice — meaning the model was essentially blind by design on debutants and deferred to market pricing. Grade those accordingly: the market's side won all three.

Two divergence flags cleared the 8-percentage-point threshold:

• Vitor Petrino vs. Serghei Spivac: model had Petrino at 73%, market at 60% — a 13.5pp gap. The model's side won. The board was right to lean harder on Petrino than the market did.
• Roman Dolidze vs. Reinier de Ridder: model had Dolidze at 36% (de Ridder at 64%), market had de Ridder at 73% — a 9.5pp gap where the model's side was Dolidze. Dolidze lost. The market read that room better.

The two misses at the winner level: Hernandez in the main event (model 58%, market 61% — both wrong) and Nzechukwu in the Gaziev fight (model had Nzechukwu at 51% — a coin flip that went the other way). No biggest_misses were flagged beyond what's already in the divergence record.

Seven bouts with results. Five finished by KO/TKO. Zero submissions. Two went to decision. That's a violent card by any measure — when the night produces a finish in five of seven, you're watching something.

The decisions: Rodrigues over Hernandez at five rounds, Petrino over Spivac at three. Everything else ended early.

One note that has to be stated plainly: the published board journals DK-point bands for each fighter — floor (p10), median, and ceiling (p90) out of the sim. Per-bout finish-method probabilities were never published. The realized methods above are results, full stop. There is no per-bout number to grade them against.

The sim projected a mean of 61.9 DK points per fighter-fight. The actual mean came in at 65.8 — a bias of -3.8 points (the model ran a little light). Of 14 bandable fighter-fights, 10 landed inside their p10–p90 band, for 71.4% coverage. The target range is 75–85%, so this card came in a tick below, driven primarily by the ceiling busts detailed in the Misses section.

Six fighters cleared 100 DK points. Here's how the board's projections held up against reality:

• Gregory Rodrigues — 135.6 actual, 61.9 projected, band 7.8–116.2. Blew past his ceiling by 19.4 points. The biggest ceiling bust on the card, and the card's top scorer.
• Reinier de Ridder — 132.0 actual, 75.4 projected, band 8.8–125.1. Above ceiling by 6.9 points. Second on the night.
• Jackson McVey — 116.0 actual, 79.3 projected, band 6.6–134.4. Landed inside the band. Clean.
• Shanelle Dyer — 107.5 actual, 88.7 projected, band 23.1–126.0. Inside the band.
• Carli Judice — 106.5 actual, 94.5 projected, band 30.2–126.5. Inside the band. The board's projection was closest to the number here.
• Shamil Gaziev — 100.8 actual, 67.2 projected, band 6.7–119.9. Inside the band.

The two fighters who landed below their floors are in the Misses section where they belong.

Own them. Here they are.

Main event wrong side: Both the model (58%) and the market (61%) had Hernandez. Rodrigues won a five-round decision. That's the loudest miss of the night — the card's marquee fight, and the board was on the wrong fighter.

Divergence flag loss: The model had Dolidze as its side in the de Ridder divergence flag (the model was lower on de Ridder than the market at 64% vs. 73%). De Ridder stopped Dolidze in the first round. The market had the better read.

Band busts — ceiling:

• Gregory Rodrigues: ceiling was 116.2, actual was 135.6 — over by 19.4 points. Biggest ceiling bust on the card per the sim.
• Reinier de Ridder: ceiling was 125.1, actual was 132.0 — over by 6.9 points.

Band busts — floor:

• Roman Dolidze: floor was 3.6, actual was 1.0 — under by 2.6 points. Biggest floor bust on the card. A first-round TKO loss will do that.
• Kennedy Nzechukwu: floor was 1.9, actual was 1.2 — under by 0.7 points. Same story: short night, bad side.

Four band misses out of 14. The ceiling busts on Rodrigues and de Ridder both trace to finish-time effects — when you go the distance in a winner's way or stop someone fast, the DK math amplifies faster than the sim's p90 captures. That's not an excuse; it's on the record.

Three things this card put on paper.

Gregory Rodrigues is a problem. Both the model and the market thought Hernandez was the better bet. Rodrigues went five rounds, took the scorecard in every judge's view, and posted the night's highest DK total at 135.6. Whatever the projections said about him going in, the result said something different.

Heavyweight is still chaos with a stopwatch. Gaziev over Nzechukwu in 80 seconds. The model had that fight essentially at a coin flip — 51% to Nzechukwu — and the market only marginally preferred Gaziev. When the data is thin and the weight class is heavyweight, the margin for error on any probability is wide. This card was a useful reminder of that.

The v0 model is 71.4% accurate on this card; the market was 85.7%. The market is still the sharper calibrator, which is exactly what you'd expect from an unvalidated first build. The Petrino divergence flag landed on the right side; the Dolidze flag did not. Two divergence flags, one win. The board is a tool in progress — that's always been the honest framing — and this card goes in the ledger as such.