The market question
The Polymarket event was straightforward: Brazil vs. Morocco, contested on 2026-06-13. The council’s graded focus was the binary-style outcome “Will Brazil win on 2026-06-13?” with related possibilities including a draw. This was framed as a World Cup opener pitting an elite favorite against a capable underdog.
What the council concluded
All four models—Gemini, Grok, Claude, and GPT—participated, with Grok acting as chairman. The council’s formal action was SKIP. It assigned a 58% probability overall and 72% confidence in that stance. Its internal most-likely outcome, however, was a Brazil win at 62%. That most-likely call became the graded pick.
The published reasoning stated: “Council recommends SKIP. Confidence: 72%. Latest previews and odds (Brazil -145 to -175) align with the 58.5% market price and historical base rates for elite favorites vs strong underdogs in World Cup openers. No breaking news or lineup shoc.” The truncated note points to the absence of late information that would have justified an edge. In short, the models saw no mispricing worth acting on and treated the favorite’s edge as already fairly reflected.
What the market was pricing
Market odds for Brazil sat in the -145 to -175 range, corresponding to roughly 58–63% implied probability. The council explicitly noted alignment with a 58.5% market price and with historical base rates for similar matchups. There was no reported divergence large enough to generate a bet; the SKIP reflected that consensus view rather than any strong conviction against the favorite.
What actually happened
The match ended in a draw. The winning outcome was “Will Brazil vs. Morocco end in a draw?” Brazil did not win. The council’s graded pick (Brazil win at 62%) was therefore incorrect. The Brier score on the graded basis was 0.3844—materially worse than a well-calibrated forecast would have produced. The council_was_correct flag is false.
This is not a case of the models inventing an exotic angle that backfired. They simply endorsed the favorite at a probability close to the market, then watched the favorite fail to convert. Draws are a known, non-trivial base rate in high-level international soccer, especially early in tournaments when both sides are cautious; the council’s reasoning referenced base rates for favorites versus strong underdogs but still concentrated probability on the win rather than fully weighting the stalemate possibility in its most-likely designation.
Takeaway
Honesty is the point of publishing every graded report. The council correctly identified that the market price looked efficient and chose to SKIP—an action that avoided a losing bet. Yet when graded on its most-likely outcome, it was wrong. Favorites do not always win openers, and a 62% lean on Brazil left too little room for the draw that actually occurred. Future soccer reports will need sharper separation between “market is efficient, skip” and “our modal outcome is still the favorite,” plus more explicit probability mass on draws when historical openers justify it. The miss is recorded, the Brier is public, and the process continues.
AI-generated analysis for informational purposes only. Not financial advice.