Judge Intelligence

Where the scores come from. Each stored card is a judge name and a bare score pair such as “27-30”. The source attaches no fighter to either number. For every bout, PropBetEdge counts which position carried the winning cards and takes that position as the winner’s — the winner necessarily won more cards than they lost. That resolves 11,821 of 12,004 cards. The remaining 183 are shown as bare pairs and marked unattributed: 180 on draws, where there is no winner to anchor on, and 3 whose tally contradicts the recorded result.

Identity. Judge names are canonicalised the same way referee names are. The upstream Details line sometimes prefixes a point deduction onto the judge’s name (“Low Blow by Watson Richard Bertrand”); those are mapped to the official and the deduction is kept as bout provenance. Merging two names is different: it combines two people’s records, so it is only done when an external judging registry holds a single official whose scored bouts account for the assignments filed under both spellings. Resemblance is not evidence, and the database refuses a merge with no source attached.

Merges applied (2). “Mamunah Querido” → “Maimunah Querido”, confirmed against MMA Decisions — judge 549 ↗ (verified 2026-09-08) by cross-matching 5 assignments. Confirms that the two archive spellings are one official. It does not settle the display spelling: the external registry renders the name "Munah Querido", which matches neither stored form, so the canonical name here stays the dominant archive spelling. “Richie Gerrard” → “Ritchie Gerard”, confirmed against MMA Decisions — judge 606 ↗ (verified 2026-09-08) by cross-matching 5 assignments. The registry's dates are US local and run a day behind ours for the Auckland and Melbourne cards, which is the dateline and not a mismatch. This pair was once merged the other way round on archive resemblance alone; the direction here comes from the source, not from which spelling the archive happens to hold more often.

What these numbers are not. A dissent count says a judge’s card differed from the official result. It does not say the card was wrong, and it is not evidence that an official favours a style, a nationality or any type of fighter — the archive cannot support that claim, so this product does not make it. Rates are withheld below 40 resolved cards and are only described as differing from the archive when the difference survives a 95% significance test against the archive baseline of 6.7%.

3 bouts in the archive have cards whose shape disagrees with the recorded method. Those are flagged on the fight page rather than corrected.