Ten minds.
One consensus.
Zero guessing.

No black box. Just receipts.

Every model gets the same match and context. We log their answers, score them after the final buzzer, and publish the leaderboard.

1. 01

   Match in

   Form, schedule, expected sets, and match context are loaded into a structured prompt for every model.

2. 02

   Parallel predictions

   Each active model forecasts scoreline, outcome, and confidence independently before kickoff.

3. 03

   Consensus + dissent

   We surface the majority pick, confidence bands, and where models disagree.

4. 04

   Score & publish

   After full-time, picks are marked against the result and the public leaderboard updates.