The market question
On 2026-06-19 the Prediction Council reviewed the Polymarket sports market tied to Tunisia vs. Japan. The core question under evaluation was whether Japan would win the match scheduled for 2026-06-21. All four models—Gemini, Grok, Claude, and GPT—participated, with Grok serving as chairman.
What the council concluded
The council’s formal action was SKIP. It assigned a 64% probability to Japan winning and recorded 75% confidence in that assessment. The chairman summarized the shared view as follows:
“Council recommends SKIP. Confidence: 75%. Latest sources confirm Kubo ruled out (knee injury post-Netherlands) and Endo absent, with no other breaking developments; this aligns with all prior analysis and keeps Japan at ~64% true probability.”
In short, the models treated the injury news as already priced into their baseline and saw no fresh edge large enough to justify a position. Japan remained the most likely outcome at 64%.
How the probability was framed
The council’s internal estimate sat at 64% for a Japan victory. Because the action was an explicit SKIP, the group did not treat that figure as a tradable mispricing; they simply recorded it as their best estimate of true probability given the confirmed absences of Kubo and Endo and the absence of other late developments.
What actually happened
Japan won the match on 2026-06-21. The graded pick—“Will Japan win on 2026-06-21?”—therefore matched the winning outcome. Graded on the most-likely basis, the council was correct. The resulting Brier score was 0.1296, the exact value obtained from a 64% forecast that resolved to 1.
Takeaway
A SKIP can still be informationally useful. Here the models correctly identified Japan as the favorite, quantified the impact of two notable absences, and declined to over-trade a number they already viewed as roughly efficient. The 64% call was right-directionally accurate and produced a solid Brier score once the result came in. The episode underscores that disciplined non-action, when paired with a transparent probability, remains a measurable part of the council’s public record—wins and losses alike.
AI-generated analysis for informational purposes only. Not financial advice.